KR20210122755A - Using context information to facilitate processing of commands in a virtual assistant - Google Patents

Abstract

The virtual assistant uses contextual information to replace natural language or gesture input from the user. Context helps elucidate the user's intent and reduces the number of candidates for interpretation of user input, and reduces the need for the user to provide excessive explanatory input. Context may include any available information that the assistant cannot substitute for explicit user input to limit information processing issues and/or to customize results. Context can be used to constrain solutions during various aspects of processing, including, for example, speech recognition, natural language processing, task flow processing, and dialog generation.

Description

USING CONTEXT INFORMATION TO FACILITATE PROCESSING OF COMMANDS IN A VIRTUAL ASSISTANT

CROSS-REFERENCE TO RELATED APPLICATIONS

This application is a continuation-in-part of U.S. Utility Application Serial No. 12/479,477 (Attorney Docket No. P7393US1), filed June 5, 2009, entitled “Contextual Voice Commands,” the entire disclosure of which is incorporated herein by reference. Claims priority as an application.

This application is a continuation-in-part of U.S. Utility Application Serial No. 12/987,982 (Attorney Docket No. P10575US1), filed January 10, 2011, entitled “Intelligent Automated Assistant,” the entire disclosure of which is incorporated herein by reference. It further claims priority as an application.

U.S. Utility Application Serial No. 12/987,982, U.S. Provisional Patent Application Serial No. 61/295,774, filed on January 18, 2010, entitled "Intelligent Automated Assistant," the entire disclosure of which is incorporated herein by reference. Claims priority from document number SIRIP003P).

[0003] This application was filed on June 3, 2011, entitled "Generating and Processing Data Items That Represent Tasks to Perform," U.S. Provisional Application Serial No. 61/493,201, the entire disclosure of which is incorporated herein by reference. It further claims priority from number P11337P1).

This application is filed on the same date as this application, entitled "Generating and Processing Task Items that Represent Tasks to Perform," the entire disclosure of which is incorporated herein by reference. / (Attorney Docket No. P11337US1).

This application was filed on the same date as this application entitled "Automatically Adapting User Interfaces for Hands-Free Interaction", the entire disclosure of which is incorporated herein by reference. / (Attorney Docket No. P11357US1).

FIELD OF THE INVENTION The present invention relates to virtual assistants and, more particularly, to mechanisms for improving the interpretation and processing of commands provided to such assistants.

Today's electronic devices are able to access a large, growing, and varying amount of functions, services and information, through the Internet and from other sources. The capabilities of such devices are increasing rapidly as many consumer devices, smartphones, tablet computers, and the like are able to run software applications to perform various tasks and to provide different types of information. Sometimes, each application, function, website or feature has its own user interface and its own operating paradigms, many of which are either cumbersome to learn or overwhelm users. Also, many users may find it difficult to even find what features and/or information is available for their electronic devices or various websites, such users may feel frustrated, overwhelmed, or simply use them. Available resources may not be used in an efficient manner.

In particular, novice users, or individuals who are not functioning normally or with disabilities, older, busy, distracted, and/or operating vehicles efficiently interface with their electronic devices and/or; It can be difficult to participate effectively in online services. In particular, such users will have difficulty with the many different and inconsistent functions, applications and websites that may be available for their use.

Thus, existing systems are sometimes difficult to use and navigate, and sometimes provide users with inconsistent and overwhelming interfaces, preventing users from efficiently utilizing the technology.

An intelligent automated assistant, also referred to herein as a virtual assistant, may provide an improved interface between a person and a computer. Implemented as described in Related U.S. Utility Application Serial No. 12/987,982 (Attorney Docket No. P10575US1), filed January 10, 2011, entitled “Intelligent Automated Assistant,” the entire disclosure of which is incorporated herein by reference. Such assistants, which may be, allow users to interact with the device or system using natural language in spoken and/or textual form. Such an assistant interprets user inputs, allows the user's intent to manipulate tasks and parameters for those tasks, executes services to support those tasks, and produces output that the user understands. .

The virtual assistant may derive any of a number of information sources to process user input including, for example, knowledge bases, models and/or data. In many cases, the user's input alone is not sufficient to clearly define the user's intent and the task to be performed. This may be due to input streams, personal differences between users, and/or ambiguities inherent in natural language. For example, a user of a text messaging application on the phone may invoke the virtual assistant and say the command "call her". Such a command is perfectly reasonable English, but it is not an accurate, workable representation, as many interpretations and possible solutions to this request may exist. Thus, without having the additional information, the virtual assistant may not be able to correctly interpret and process such input. This type of ambiguity can lead to overburdening the user with errors, performing incorrect actions, and/or requests to clarify input.

According to various embodiments of the present invention, the virtual assistant uses context information (also referred to herein as “context”) to supplement natural language or gestural input from the user. This helps clarify the user's intent, reduces the number of candidate interpretations of the user input, and reduces the need for the user to provide excessive clarification input. Context may include any available information that may be used by the assistant to supplement explicit user input to limit information-processing issues and/or personalize results. For example, if the input from the user includes a pronoun (such as "her" in "the command to call her"), the virtual assistant uses the context to refer to that pronoun. can infer, for example, the identity of the person to be called and/or the phone number to use. Other uses of context are described herein.

According to various embodiments of the present invention, any number of mechanisms may be implemented to obtain and apply context information for performing computations in a virtual assistant implemented on an electronic device. In various embodiments, the virtual assistant is filed on January 10, 2011, for the title “Intelligent Automated Assistant,” in U.S. Utility Application Serial No. 12/987,982 (Attorney Docket No. P10575US1), the entire disclosure of which is incorporated herein by reference. ) as an intelligent automated assistant as described in Such an assistant establishes a relationship with the user in an integrated conversational manner using natural language conversation, and calls external services when it is appropriate to obtain information or perform various operations. According to the techniques described herein, contextual information is used in such an assistant to reduce ambiguity when performing information processing functions such as, for example, speech recognition, natural language processing, task flow processing, and dialog generation.

According to various embodiments of the present invention, a virtual assistant uses context to perform various and different kinds of operations, functions and/or features, and/or a plurality of features of an electronic device on which it is installed. may be configured, designed, and/or operable to combine operations, operations and applications. In some embodiments, the virtual assistant of the present invention actively elicits input from the user, interprets the user's intent, disambiguates contention interpretations, requests and receives explicit information as needed. Context may be used when performing either or both of performing (or initiating) actions based on the action and/or the identified intent.

For example, operations may be performed by activating and/or interfacing not only applications or services available on the electronic device, but also services available through an electronic network such as the Internet. In various embodiments, activation of such external services may be performed by application programming interfaces (APIs) or some other suitable mechanism. As such, the virtual assistant implemented in accordance with various embodiments of the present invention integrates, simplifies, and improves the user's experience for many different applications and functions of the electronic device and for services available over the Internet. can do. Thereby, the user is less burdened with learning what features are available on the device and in web-connected services, how to interface with those services to get what they want, and how to interpret the output they receive from those services. can Rather, the virtual assistant of the present invention may act as an intermediary between the user and such various services.

Further, in various embodiments, the virtual assistant of the present invention provides an interactive interface that the user may find more easy to use and less burdensome than a conventional graphical user interface. The user establishes a relationship in the form of a conversation used in daily conversation with the virtual assistant using any one of a number of available input and output mechanisms, such as, for example, voice, graphical user interfaces (buttons and links), text entry, and the like. can tie The system may be implemented using any of a number of different platforms, such as device APIs, web, email, and the like. Requests for additional input may be presented to the user in the context of such a conversation. Short-term and long-term memory can be used to interpret user input into an appropriate context, taking into account historical and profile information about the user as well as previous events and communications within a given session.

Further, in various embodiments, contextual information obtained from a user's interaction with a feature, operation, or application on a device may be used to simplify the operation of other features, operations, or applications on that device or other device. . For example, a virtual assistant may use the context of a phone call (such as the person calling) to streamline the initiation of a text message (e.g., a text message should be sent to the same person without the user explicitly specifying the recipient of the text message). can be decided). As such, the virtual assistant of the present invention can interpret commands such as "send him a text message", where "he" is contextual information obtained from the current phone call and/or from any feature, action or application on the device. interpreted according to In various embodiments, the virtual assistant considers various types of available context data to determine which address book contacts to use, which contact data to use, which phone numbers to use for contacts, and the like, so that the user can Eliminate the need to manually reassign such information.

The contextual information source may include the current time, location, current state of the device used as an interface to the virtual assistant, such as an application or data object; personal data such as the user's address book, calendar and application usage history; and states of conversations between the user and the virtual assistant, such as recently mentioned people and/or places, which are exemplary and not limiting.

Context can be applied to various calculations and inferences in the operation of the virtual assistant. For example, context can be used to reduce ambiguity or constrain the number of solutions when processing user input. Accordingly, the context may be used to constrain solutions during various processing steps, which are illustrative and not restrictive.

• Speech Recognition - Receive voice input and generate candidate interpretations as text, such as "call her", "collar" and "call Herb". The context may be used to constrain which words and phrases to consider by the speech recognition module, how to rank them, and which ones to accept above a threshold for consideration. For example, a user's address book adds personal names to other language-generic models of speech, allowing these names to be recognized and prioritized.

Natural Language Processing (NLP) - parses text and associates words with syntactic and semantic rules, such as determining that user input relates to making a phone call to the person represented by the pronoun "she", To find a specific data representation. For example, the context of a text messaging application can help constrain the interpretation of "her" to mean "the person I am texting with."

• Task Flow Processing —Identifies the user task, task steps, and task parameters used to assist in that task, such as which phone number to use for the person "she" represents. Additionally, the context of the text messaging application may constrain the interpretation of the phone number to indicate that the system should use the current or recently used number for text messaging conversations.

Create Conversation - Generate assistant responses as part of the conversation with the user about their task, eg translating the user's intent into a response "Yeah, I'll call Rebecca on her mobile..." (paraphrase) ). Levels of verbosity and informal tone are options that can be guided by contextual information.

In various embodiments, the virtual assistant of the present invention may control various features and operations of the electronic device. For example, the virtual assistant may call services that interface with functions and applications on the device by APIs or other means to perform functions and operations that should have been initiated using a conventional user interface on the device. Such functions and actions may include, for example, setting an alarm, making a call, sending a text message or email message, adding a calendar event, and the like. Such functions and operations may be performed as additional functions in the context of a conversation used in a daily conversation between a user and a virtual assistant. Such functions and actions may be specified by a user in the context of such a conversation, or may be performed automatically based on the context of that conversation. Those of ordinary skill in the art will appreciate that it can thereby be used as a control mechanism for initiating and controlling various operations on an electronic device, and thus as an alternative to conventional mechanisms such as buttons or graphical user interfaces. As described herein, contextual information may be used to inform and enhance the use of such virtual assistants as control mechanisms.

BRIEF DESCRIPTION OF THE DRAWINGS The accompanying drawings illustrate several embodiments of the invention, and together with the description, serve to explain the principles of the invention according to the embodiments. Those skilled in the art will recognize that the specific embodiments shown in the drawings are illustrative only and are not intended to limit the scope of the invention.
1 is a block diagram illustrating some examples of a virtual assistant and sources of context that may affect its operation, in accordance with one embodiment.
2 is a flow diagram illustrating a method of using context in various stages of processing in a virtual assistant, according to one embodiment.
3 is a flow diagram illustrating a method of using context in speech elicitation and interpretation, according to one embodiment.
4 is a flow diagram illustrating a method of using context in natural language processing, according to one embodiment.
5 is a flowchart illustrating a method of using context in task flow processing, according to one embodiment.
6 is a block diagram illustrating an example of sources of context distributed between a client and a server, according to one embodiment.
7A-7D are event diagrams illustrating examples of mechanisms for obtaining and manipulating context information in accordance with various embodiments;
8A-8D illustrate examples of various representations of context information that may be used in connection with various embodiments of the present invention;
9 is a diagram illustrating an example of a settings table specifying communication and caching policies for various contextual information sources, according to one embodiment.
FIG. 10 is an event diagram illustrating an example of accessing the contextual information sources set up in FIG. 9 during processing of an interaction sequence, according to one embodiment;
11-13 are a series of screenshots illustrating examples of using an application context in a text messaging domain to derive referents to pronouns, according to one embodiment.
14 is a screenshot illustrating a virtual assistant prompting for name disambiguation, according to one embodiment.
15 is a screenshot illustrating a virtual assistant using a conversation context to infer a location for a command, according to one embodiment.
16 is a screenshot illustrating an example of using a phone preference list as a source of context, according to one embodiment.
17-20 are a series of screen shots illustrating an example of using a current application context to interpret and operate a command, according to one embodiment.
21 is a screenshot showing an example of using the current application context to interpret commands that invoke different applications.
22-24 are a series of screenshots illustrating examples of using event context in the form of an incoming text message, according to one embodiment.
25A and 25B are a series of screen shots illustrating examples of using previous conversation text, according to one embodiment.
26A and 26B are screen shots illustrating an example of a user interface for selecting one of candidate interpretations, according to one embodiment.
27 is a block diagram illustrating an example of one embodiment of a virtual assistant system.
28 is a block diagram illustrating a computing device suitable for implementing at least a portion of a virtual assistant in accordance with at least one embodiment.
29 is a block diagram illustrating an architecture for implementing at least a portion of a virtual assistant on a standalone computing system, according to at least one embodiment.
30 is a block diagram illustrating an architecture for implementing at least a portion of a virtual assistant on a distributed computing network, according to at least one embodiment.
31 is a block diagram illustrating a system architecture illustrating several different types of clients and modes of operation.
32 is a block diagram illustrating a client and a server communicating with each other to implement the present invention in accordance with one embodiment.

According to various embodiments of the present invention, various context information is obtained and applied to perform information processing functions in support of operations of the virtual assistant. For purposes of explanation, the term “virtual assistant” is equivalent to the term “intelligent automated assistant”, both referring to any information processing system that performs one or more of the functions below.

음성(spoken) 또는 텍스트 형태로, 인간 언어 입력을 해석

Interpret human language input in spoken or text form

단계들 및/또는 파라미터들로 태스크를 표현하는 것과 같은, 실행될 수 있는 형태로 사용자 의도를 표현하는 것의 동작화.

Operation of expressing user intent in an executable form, such as expressing a task with steps and/or parameters.

프로그램들, 메소드들, 서비스들, API들 등을 호출함으로써 태스크 표현들을 실행; 및

executing task expressions by calling programs, methods, services, APIs, etc.; and

언어 및/또는 그래픽 형태로 사용자에 대한 출력 응답들을 생성.

Generate output responses to the user in verbal and/or graphical form.

An example of such a virtual assistant is disclosed in related U.S. Utility Model No. 12/987,982, entitled “Intelligent Automated Assistant” (reference number P10575US1), filed January 10, 2011, the entire disclosure of which is incorporated herein by reference. Included.

Various techniques will now be described in detail with reference to embodiments as shown in the accompanying drawings. In the following description, numerous specific details are set forth in order to provide a thorough understanding of one or more aspects and/or features or reference herein described. It will be apparent, however, to one skilled in the art, that one or more aspects and/or features or reference described herein may be practiced without some or all of these specific details. In other instances, well-known process steps and/or structures have not been described in detail in order not to obscure some of the aspects and/or features or references described herein.

One or more different inventions may be described in this application. In addition, various embodiments of one or more invention(s) described herein may be set forth in this patent application and are presented for illustrative purposes only. The described embodiments are not intended to be limiting in any sense. One or more of the invention(s) may be broadly applicable to various embodiments, as will be apparent from the disclosure. These embodiments are described in sufficient detail to enable those skilled in the art to practice one or more of the invention(s), and other embodiments may be utilized, and structural, logical, software, electrical and other modifications may be made. It will be understood that these may be made without departing from the scope of the invention(s). Accordingly, those skilled in the art will recognize that one or more of the invention(s) may be practiced with various modifications and alternatives. One or more specific features of the invention(s) form part of the present disclosure or one or more specific embodiments, and described with reference to the drawings in which one or more specific embodiments of the invention(s) are shown by way of illustration. can be It should be understood, however, that such features are not limited to use in one or more specific embodiments or figures to which they are described with reference. This disclosure is neither a literal description of all embodiments of one or more of the invention(s) nor a listing of features of one or more of the invention(s) that should be present in all embodiments.

The headings of the sections provided in this patent application and the headings of this patent application are for convenience only and should not be taken as limiting the present disclosure in any way.

Devices communicating with each other need not be in continuous communication with each other, unless expressly specified otherwise. Also, devices that communicate with each other may communicate directly or indirectly through one or more intermediaries.

A description of embodiments using several components in communication with each other does not imply that all such components are required. Conversely, various optional components are described to illustrate a wide range of possible embodiments of one or more of the invention(s).

Further, although process steps, method steps, algorithms, etc. may be described in a sequential order, such processes, methods, and algorithms may be configured to operate in any suitable order. In other words, any sequence or order of steps that may be described in this patent application does not in itself represent a requirement that the steps be performed in that order. Also, while described or suggested to occur asynchronously (eg, because one step is described after another), some steps may be performed concurrently. Moreover, the illustration of a process by its depiction in the drawings does not mean that the illustrated process excludes other modifications and variations thereto, and that the illustrated process or any of its steps is not intended to be subject to the present invention(s). ), and does not imply that the illustrated process is preferred.

Where a single device or article is described, more than one device/article (whether they cooperate or not) may be used in place of the single device/article. Likewise, where more than one device or article is described (whether they collaborate or not), it will be apparent that a single device/article may be used in place of more than one device or article.

The functionality and/or features of a device may alternatively be implemented by one or more other devices not explicitly described as having such functionality/features. Accordingly, other embodiments of one or more of the invention(s) do not require the device itself.

Techniques and mechanisms described or referenced herein will at times be described in singular form for purposes of clarity. It should be noted, however, that certain embodiments include multiple exemplifications of mechanisms or iterations of multiple techniques, unless stated otherwise.

Although described within the context of technology for implementing an intelligent automated assistant, also known as a virtual assistant, various aspects and techniques described herein may also be employed and/or computerized with humans and/or software It can be applied to other fields of technology that involve interactivity.

Virtual assistant technology (eg, utilized, provided, and/or implemented by one or more virtual assistant system embodiments described herein) is disclosed in one or more below, the entire disclosure of which is herein The specification is incorporated by reference.

미국 실용 출원 번호 제12/987,982호, "Intelligent Automated Assistant", 대리인 정리 번호 P10575US1, 2011.1.10 제출.

U.S. Utility Application No. 12/987,982, "Intelligent Automated Assistant", filed at agent's docket number P10575US1, 10 January 2011.

미국 가특허 출원 번호 제61/295,774호, "Intelligent Automated Assistant", 대리인 정리 번호 SIRIP003P, 2010.1.18 제출.

U.S. Provisional Patent Application No. 61/295,774, "Intelligent Automated Assistant," filed at Attorney's Office No. SIRIP003P, January 18, 2010.

미국 특허 출원 번호 제11/518,292호, "Method And Apparatus for Building an Intelligent Automated Assistant", 2006년 9월 8일 제출.

U.S. Patent Application No. 11/518,292, "Method And Apparatus for Building an Intelligent Automated Assistant," filed September 8, 2006.

미국 가특허 출원 번호 제61/186,414호, "System and Method for Semantic Auto-Completion", 2009.6.12 제출.

U.S. Provisional Patent Application No. 61/186,414, "System and Method for Semantic Auto-Completion," filed June 12, 2009.

hardware architecture

In general, the virtual assistant techniques disclosed herein may be implemented on hardware or a combination of software and hardware. For example, they may be implemented as an operating system kernel on a specially configured machine and/or on a network interface card, as a separate user process, as a library bound to network applications. In certain embodiments, the techniques disclosed herein may be implemented in software, such as an operating system or an application running on an operating system.

The software/hardware hybrid implementation(s) of at least some of the virtual assistant embodiment(s) disclosed herein may be implemented on a programmable machine selectively activated or reconfigured by a computer program stored in memory. Such network devices may have multiple network interfaces that may be configured or designed to use different types of network communication protocols. The general architecture for some of these machines may be apparent from the description disclosed herein. According to certain embodiments, at least some of the features and/or functionalities of the various virtual assistant embodiments disclosed herein may include one or more general purpose network host machines, such as an end user computer system, computer, network server or server system; a mobile computing device (e.g., personal digital assistant (PDA), mobile phone, smartphone, laptop, tablet computer, etc.), consumer electronic device, music player or any other suitable electronic device such as a router, switch, or the like, or a combination thereof. can be implemented. In at least some embodiments, at least some of the features and/or functions of the various virtual assistant embodiments disclosed herein may be implemented in one or more virtualized computing environments (eg, network computing clouds, etc.). .

Referring now to FIG. 28 , shown is a block diagram illustrating a computing device 60 suitable for implementing at least some of the virtual assistant features and/or functions disclosed herein. Computing device 60 may be, for example, an end user computer system, a network server or server system, a mobile computing device (eg, personal digital assistant (PDA), mobile phone, smartphone, laptop, tablet computer, etc.) , a consumer electronics device, a music player, or any other suitable electronic device, or any combination or portion thereof. Computing device 60 may be configured to communicate wirelessly or wired with other computing devices, such as clients and/or servers, via a communication network such as the Internet, using known protocols for such communication.

In one embodiment, computing device 60 includes a central processing unit (CPU) 62 , interfaces 68 , and a bus 67 (such as peripheral component interconnect (PCI)). When operating under the control of appropriate software or firmware, CPU 62 may be responsible for implementing particular functions associated with the functions of a specifically configured computing device or machine. For example, in at least one embodiment, the user's personal digital assistant (PDA) or smartphone functions as a virtual assistant system utilizing CPU 62 , memory 61 , 65 , and interface(s) 68 . may be constructed or designed to In at least one embodiment, CPU 62 has one or more different types of software modules/components under the control of which may include, for example, an operating system and any suitable application software, drivers, etc. may be brought about to perform virtual assistant functions and/or tasks of

CPU 62 may include one or more processor(s) 63 such as, for example, a Motorola processor or an Intel microprocessor family or a MIPS microprocessor family. In some embodiments, the processor(s) 63 include hardware specifically designed to control the operations of the computing device 60 (eg, application-specific integrated circuits (ASICs), electrically erasable programmable read- only memories), field-programmable gate arrays (FPGAs), etc.). In certain embodiments, memory 61 (such as non-volatile random access memory (RAM) and/or read-only memory (ROM)) also forms part of CPU 62 . However, there are many different ways in which a memory may couple with a system. The memory block 61 may be used for a variety of purposes, such as, for example, caching and/or storage of data, programming instructions, and the like.

As used herein, the term "processor" is not limited to just those integrated circuits referred to in the art as a processor, but is not limited to a microcontroller, a microcomputer, a programmable logic controller, an ASIC. (application-specific integrated circuit), and any other programmable circuit.

In one embodiment, interfaces 68 are provided as interface cards (sometimes referred to as “line cards”). In general, they control the transmission and reception of data packets over a computing network and support other peripherals that are sometimes used with computing device 60 . Among the interfaces that may be provided are Ethernet interfaces, frame relay interfaces, cable interfaces, DSL interfaces, token ring interfaces, and the like. Also, e.g. Universal Serial Bus (USB), Serial, Ethernet, FireWire, PCI, Parallel, Radio Frequency (RF), Bluetooth ^TM , Near Field Communication (e.g. using Near Field Magnetics), 802.11 (WiFi ), Frame Relay, TCP/IP, ISDN, Fast Ethernet interfaces, Gigabit Ethernet interfaces, Asynchronous Transfer Mode (ATM) interfaces, High Speed Serial Interface (HSSI) interfaces, Point of Sale (POS) interfaces, FDDI ( Various types of interfaces may be provided, such as fiber data distributed interfaces). In general, these interfaces 68 may include ports suitable for communication with an appropriate medium. In some cases, they may also include an independent processor, and in some instances may include volatile and/or non-volatile memory (eg, RAM).

Although the system shown in FIG. 28 depicts one particular architecture for a computing device 60 for implementing the inventive techniques described herein, it is by no means implemented that at least some of the features and techniques described herein are implemented. It's not the only device architecture that can be. For example, architectures having one or any number of processors 63 may be used, and such processors 63 may be present within one device or distributed among any number of devices. In one embodiment, one processor 63 handles communications along with routing calculations. In various embodiments, different types of virtual assistant features and/or functions may be combined with a client device (such as a personal digital assistant or smartphone running client software) and a server system (such as a server system to be described in greater detail below). may be implemented in a virtual assistant system comprising (s).

Irrespective of the network device configuration, the system of the present invention may include one or more memories or (e.g., For example, memory modules (such as memory block 65) may be employed. Program instructions may, for example, control operation of an operating system and/or one or more application programs. The memory or memories may also be configured to store data structures, keyword classification information, advertisement information, user click and impression information, and/or other specific non-program information described herein.

Because such information and program instructions may be employed to implement the systems/methods described herein, at least some network device embodiments may include a nontransitory machine-readable storage medium, A temporary machine-readable storage medium may be configured or designed to store, for example, program instructions, state information, and the like for performing various tasks described herein. Examples of such non-transitory machine-readable media include magnetic media such as hard disks, floppy disks, and magnetic tape; optical media such as CD-ROM disks; magneto-optical media, such as floptical disks, and read-only memory devices (ROM) specially configured to store and execute program instructions, flash memory, memristor memory, random access memory (RAM) hardware devices such as, but not limited to. Examples of program instructions include both machine code, such as generated by a compiler, and files containing higher-level code that can be executed by a computer using an interpreter.

In one embodiment, the system of the present invention is implemented on a standalone computing system. Referring now to FIG. 29 , shown is a block diagram illustrating an architecture for implementing at least a portion of a virtual assistant on a standalone computing system in accordance with at least one embodiment. Computing device 60 includes process(es) 63 executing software implementing virtual assistant 1002 . The input device 1206 may be, for example, a keyboard, a touchscreen, a microphone (eg, for voice input), a mouse, a touchpad, a trackball, a five-way switch, a joystick, and/or their It can be of any type suitable for receiving user input, including binding. The output device 1207 may be a screen, a speaker, a printer, and/or a combination thereof. Memory 1210 may be a random access memory having a structure and architecture as known in the art, used by processor(s) 63 while executing software. Storage device 1208 may be any magnetic, optical, and/or electrical storage device for storing data in digital form; Examples include flash memory, magnetic hard disks, CD-ROMs, and the like.

In another embodiment, the system of the present invention is implemented on a distributed computing network, such as having any number of clients and/or servers. Referring now to FIG. 30 , shown is a block diagram illustrating an architecture for implementing at least a portion of a virtual assistant on a distributed computing network, according to at least one embodiment.

In the arrangement shown in FIG. 30 , any number of clients 1304 are provided; Each client 1304 may execute software for implementing the client-side portions of the present invention. Also, any number of servers 1340 may be provided to handle requests received from clients 1304 . Clients 1304 and servers 1340 may communicate with each other via an electronic network 1361 , such as the Internet. Network 1361 may be implemented using any known network protocols including, for example, wired and/or wireless protocols.

Also, in one embodiment, servers 1340 may call external services 1360 when it is necessary to obtain additional information or store data regarding previous interactions with particular users. Communication with external services 1360 may occur, for example, via network 1361 . In various embodiments, external services 1360 include web-enabled services and/or functionality related to or installed on the hardware device itself. For example, in embodiments where assistant 1002 is implemented in a smartphone or other electronic device, assistant 1002 obtains information stored in a calendar application (“app”), contacts, and/or other sources. can do.

In various embodiments, assistant 1002 may control multiple features and operations of the electronic device on which it is installed. For example, assistant 1002 may interface with the functionality and applications of the device via APIs or by other means to perform functions and operations that could otherwise be initiated using a conventional user interface of the device. External services 1360 may be called. These functions and actions may include, for example, setting an alarm, making a call, sending a text message or email message, adding a calendar event, and the like. These functions and operations may be performed as additional functions in the context of a conversation used in a daily conversation between the user and assistant 1002 . These functions and actions may be specified by the user in the context of such a conversation, or may be performed automatically based on the context of the conversation. Those skilled in the art will appreciate that assistant 1002 may thereby be used as a control mechanism for initiating and controlling various operations of an electronic device, and that the electronic device may be used as an alternative to conventional mechanisms such as buttons or graphical user interfaces. will recognize that

For example, the user may provide the input to assistant 1002 "I need to get up at 8 o'clock tomorrow morning". Once assistant 1002 determines the user's intent using the techniques described herein, assistant 1002 calls external services 1340 to interface with the device's alarm clock function or application. The assistant 1002 sets an alarm on behalf of the user. In this way, a user may use assistant 1002 as a replacement for conventional mechanisms for setting an alarm or performing other functions of the device. If the user's requests are ambiguous or require further clarification, assistant 1002 can provide active elicitation, paraphrasing, suggestions, etc., as described herein, including acquisition context information. Various techniques may be used so that the correct services 1340 are invoked and the intended action taken. In one embodiment, assistant 1002 may prompt the user to request and/or confirm additional contextual information from any suitable source prior to invoking service 1340 to perform a function. In one embodiment, the user may selectively disable assistant 1002's ability to invoke certain services 1340, or may disable all such service calls if desired.

The system of the present invention may be implemented using any of a number of different types of operating modes and clients 1304 . Referring now to FIG. 31 , a block diagram depicting a system architecture illustrating several different types of clients 1304 and modes of operation is shown. One of ordinary skill in the art will appreciate that the various types of clients 1304 and the modes of operation shown in FIG. 31 are merely exemplary, and the system of the present invention may utilize modes of operation and/or clients 1304 other than those shown. It will be appreciated that it can be implemented. Further, the system may include any or all of the modes of operation and/or such clients 1304 , alone or in any combination. Examples shown include:

입력/출력 디바이스들 및/또는 센서들(1402)을 갖는 컴퓨터 디바이스들. 클라이언트 컴포넌트는 임의의 이러한 컴퓨터 디바이스(1402)에 배치될 수 있다. 적어도 하나의 실시예는 네트워크(1361)를 통한 서버들(1340)과의 통신을 가능하게 하기 위해 웹 브라우저(1304A) 또는 다른 소프트웨어 애플리케이션을 사용하여 구현될 수 있다. 입력 및 출력 채널들은 예를 들어 시각 및/또는 청각 채널들을 포함하는 임의의 유형일 수 있다. 예를 들어, 일 실시예에서, 본 발명의 시스템은, 웹 브라우저의 동등물이 음성에 의해 구동되고 출력을 위해 음성을 사용하는 맹인용 비서의 실시예를 가능하게 하는, 음성 기반 통신 방법들을 사용하여 구현될 수 있다.

Computer devices with input/output devices and/or sensors 1402 . The client component may be located on any such computer device 1402 . At least one embodiment may be implemented using a web browser 1304A or other software application to enable communication with servers 1340 over network 1361 . The input and output channels may be of any type, including, for example, visual and/or auditory channels. For example, in one embodiment, the system of the present invention uses voice-based communication methods to enable an embodiment of an assistant for the blind whose equivalent of a web browser is voice-driven and uses voice for output. can be implemented.

클라이언트가 모바일 디바이스(1304B)의 애플리케이션으로서 구현될 수 있는 I/O 및 센서들(1406)을 갖는 모바일 디바이스들. 이것은 휴대폰, 스마트폰, PDA, 태블릿 디바이스, 네트워크 게임 콘솔 등을 포함하지만 이에 한정되는 것은 아니다.

Mobile devices with I/O and sensors 1406 in which the client may be implemented as an application of the mobile device 1304B. This includes, but is not limited to, cell phones, smart phones, PDAs, tablet devices, network game consoles, and the like.

클라이언트가 기기(1304C)의 임베딩된 애플리케이션으로서 구현될 수 있는 I/O 및 센서들(1410)을 갖는 소비자 가전들.

Consumer appliances with sensors 1410 and I/O where the client may be implemented as an embedded application of device 1304C.

클라이언트가 임베딩된 시스템 애플리케이션(1304D)으로서 구현될 수 있는 대시보드 인터페이스들 및 센서들(1414)을 갖는 승용차들 및 다른 차량들. 이것은 네비게이션 시스템, 음성 제어 시스템, 차량내 엔터테인먼트 시스템 등을 포함하지만 이에 한정되는 것은 아니다.

Cars and other vehicles with dashboard interfaces and sensors 1414 in which the client can be implemented as an embedded system application 1304D. This includes, but is not limited to, navigation systems, voice control systems, in-vehicle entertainment systems, and the like.

클라이언트가 디바이스 상주 애플리케이션(1304E)으로서 구현될 수 있는, 라우터들(1418) 또는 네트워크에 상주하거나 네트워크와 인터페이스하는 임의의 다른 디바이스 등의 네트워킹된 컴퓨팅 디바이스들.

Networked computing devices, such as routers 1418 or any other device that resides in or interfaces with a network, where the client may be implemented as a device resident application 1304E.

비서의 실시예가 이메일 모달리티 서버(1426)를 통해 접속되는 이메일 클라이언트들(1424). 이메일 모달리티 서버(1426)는, 예를 들어 사용자로부터의 입력을 비서로 전송된 이메일 메시지들로서 수취하고 응답들로서 비서로부터의 출력을 사용자에게 전송하는, 통신 브리지로서 동작한다.

Email clients 1424 to which an embodiment of the assistant is connected via an email modality server 1426. Email modality server 1426 acts as a communication bridge, for example, receiving input from the user as email messages sent to the assistant and sending output from the assistant to the user as responses.

비서의 실시예가 메시지 모달리티 서버(1430)를 통해 접속되는 인스턴트 메시지 클라이언트들(1428). 메시지 모달리티 서버(1430)는, 사용자로부터의 입력을 비서로 전송된 메시지들로서 수취하고 비서로부터의 출력을 응답 시의 메시지들로서 사용자에게 전송하는, 통신 브리지로서 동작한다.

Instant message clients 1428 to which an embodiment of the assistant is connected via a message modality server 1430. Message modality server 1430 acts as a communication bridge, receiving input from the user as messages sent to the assistant and sending output from the assistant to the user as messages in response.

비서의 실시예가 VoIP(Voice over Internet Protocol) 모달리티 서버(1430)를 통해 접속되는 음성 전화들(1432). VoIP 모달리티 서버(1430)는, 사용자로부터의 입력을 비서에게 말한 음성으로서 수취하고 비서로부터의 출력을 예를 들어 응답 시의 합성된 음성으로서 사용자에게 전송하는, 통신 브리지로서 동작한다.

Voice calls 1432 to which an embodiment of the assistant is connected via a Voice over Internet Protocol (VoIP) modality server 1430 . VoIP modality server 1430 acts as a communication bridge, receiving input from the user as voice spoken to the secretary and sending output from the assistant to the user, for example, as synthesized voice in response.

For messaging platforms including, but not limited to, email, instant messaging, discussion forums, group chat sessions, real-time help or customer support sessions, etc., assistant 1002 may function as a participant in conversations. have. Assistant 1002 may monitor the conversation and respond to individuals or groups using one or more techniques and methods disclosed herein for one-on-one interactions.

In various embodiments, the functionality for implementing the techniques of the present invention may be distributed among any number of client and/or server components. For example, various software modules may be implemented to perform various functions related to the present invention, and such modules may be variously implemented to run on server and/or client components. Additional details of such a configuration can be found in related U.S. Patent Application Serial No. 12/987,982, “Intelligent Automated Assistant,” Attorney Docket No. P10575US1, filed Jan. 10, 2011, which is incorporated herein by reference in its entirety. is provided

In the example of FIG. 32 , the input derivation function and the output processing function are distributed to the client 1304 and the server 1340 , the client portion 2794a of the input derivation and the client portion 2792a of the output processing are distributed to the client 1304 . The server portion 2794b of input derivation and the server portion 2792b of output processing are located on the server 1340 . The following components are located in server 1340:

완성된 어휘(2758b);

completed vocabulary (2758b);

언어 패턴 인식기들의 완성된 라이브러리(2760b);

a complete library of language pattern recognizers 2760b;

단기 개인 메모리의 마스터 버전(2752b);

master version of short-term personal memory (2752b);

장기 개인 메모리의 마스터 버전(2754b).

Master version of long-term private memory (2754b).

In one embodiment, the client 1304 maintains subsets and/or portions of these components locally to improve responsiveness and reduce reliance on network communications. Such subsets and/or portions may be maintained and updated according to known cache management techniques. Such subsets and/or portions include, for example:

어휘의 서브세트(2758a);

a subset of vocabulary 2758a;

언어 패턴 인식기들의 라이브러리의 서브세트(2760a);

a subset of the library of language pattern recognizers 2760a;

단기 개인 메모리의 캐시(2752a);

a cache of short-term personal memory 2752a;

장기 개인 메모리의 캐시(2754a).

Cache of long-term private memory (2754a).

Additional components may be implemented as part of server 1340 including, for example, the following.

언어 해석기(2770);

language interpreter 2770;

대화 플로우 프로세서(2780);

conversation flow processor 2780;

출력 프로세서(2790);

output processor 2790;

도메인 엔티티 데이터베이스들(2772);

domain entity databases 2772;

태스크 플로우 모델들(2786);

task flow models 2786;

서비스 편성(2782);

service organization 2782;

서비스 능력 모델들(2788).

Service capability models 2788.

Each of these components will be described in more detail below. Server 1340 obtains additional information when needed by interfacing with external services 1360 .

conceptual architecture

Referring now to FIG. 27 , a simplified block diagram of a specific example embodiment of virtual assistant 1002 is shown. As disclosed in more detail in the aforementioned related US patent applications, different embodiments of virtual assistant 1002 may implement a variety of different types of operations, functions, and/or features generally related to virtual assistant technology. may be constructed, designed, and/or operable to provide Further, as disclosed in greater detail herein, many of the various operations, functions, and/or features of virtual assistant 1002 disclosed herein are different for different entities interacting with virtual assistant 1002 . may enable or provide tangible advantages and/or advantages. The embodiment shown in FIG. 27 may be implemented using any of the hardware architectures described above, or using a different type of hardware architecture.

For example, according to different embodiments, virtual assistant 1002 may implement various different types of operations, functionalities, and/or features, such as, for example, one or more of the following (or combinations thereof): may be constructed, designed, and/or operable to provide:

제품들 및 서비스들을 찾거나, 발견하거나, 그 중에서 선택하거나, 구매하거나, 예약하거나, 또는 주문하기 위해 인터넷을 통해 이용가능한 데이터 및 서비스들의 애플리케이션을 자동화함. 이들 데이터 및 서비스들을 이용하는 것의 프로세스를 자동화하는 것에 더하여, 가상 비서(1002)는 또한 한번에 데이터 및 서비스들의 몇몇 소스들의 결합된 사용을 가능하게 할 수 있다. 예를 들면, 그것은 몇몇 리뷰 사이트들로부터 제품들에 관한 정보를 결합하고, 다수의 유통업자로부터 가격 및 입수가능성을 확인하고, 그들의 위치들 및 시간 제약들을 확인하고, 사용자가 그들의 문제에 대한 개인화된 해법을 찾는 것을 도와줄 수 있다.

Automating the application of data and services available over the Internet to find, discover, select from, purchase, book, or order products and services. In addition to automating the process of using these data and services, virtual assistant 1002 may also enable the combined use of several sources of data and services at once. For example, it combines information about products from several review sites, checks prices and availability from multiple distributors, checks their locations and time constraints, and lets users personalize their concerns about their issues. We can help you find a solution.

(영화들, 이벤트들, 공연들, 전시들, 쇼들 및 어트랙션들을 포함하나 이에 제한되지 않는) 해야할 일들; (여행 목적지들, 호텔들 및 머무르기 위한 그외의 장소들, 랜드마크들 및 관심 있는 그외의 사이트들, 등을 포함하나 이에 제한되지 않는) 가야할 장소들; (레스토랑 및 바와 같은) 먹거나 또는 마시기 위한 장소들, 다른 사람들을 만나기 위한 시간들 및 장소들, 및 인터넷 상에서 찾을 수 있는 엔터테인먼트 또는 소셜 상호작용의 임의의 그외의 소스들에 대해 찾거나, 조사하거나, 그 중에서 선택하거나, 예약하거나, 그렇지 않은 경우 배우기 위해 인터넷을 통해 이용가능한 데이터 및 서비스들의 이용을 자동화함.

things to do (including but not limited to movies, events, performances, exhibits, shows and attractions); places to go (including but not limited to travel destinations, hotels and other places to stay, landmarks and other sites of interest, etc.); finding, researching, or searching for places to eat or drink (such as restaurants and bars), times and places to meet other people, and any other sources of entertainment or social interaction that can be found on the Internet; Automating the use of data and services available over the Internet to select from, book, or otherwise learn.

(위치 기반 탐색을 포함하는) 탐색; 네비게이션(지도들 및 방향들); (이름 또는 그외의 속성들에 의해 비지니스들 또는 사람들을 찾는 것과 같은) 데이터베이스 검색; 기상 상태 및 예보를 얻는 것, 마켓 아이템들의 가격 또는 금융 거래의 상태를 확인하는 것; 교통 상황 또는 항공편의 상태를 모니터링하는 것; 캘린더들 및 스케쥴들을 액세스하고 업데이트하는 것; 리마인더들, 경보들, 태스크들, 및 프로젝트들을 관리하는 것; 이메일 또는 그외의 메시징 플랫폼들을 통해 통신하는 것; 및 국부적으로 또는 원격으로 디바이스들을 동작시키는 것(예를 들면, 전화를 거는 것, 전등 및 온도를 제어하는 것, 가정 보안 디바이스들을 제어하는 것, 음악 또는 비디오를 플레이하는 것, 등)을 포함하는 그래픽 사용자 인터페이스들을 갖는 전용 애플리케이션들에 의해 다르게 제공되는 자연어 대화를 통해 애플리케이션들 및 서비스들의 동작을 가능하게 함. 일 실시예에서, 가상 비서(1002)는 디바이스 상에서 이용가능한 많은 기능들 및 앱들을 시작하고, 동작시키고, 제어하는 데에 이용될 수 있다.

navigation (including location-based navigation); navigation (maps and directions); database search (such as finding businesses or people by name or other attributes); obtaining weather conditions and forecasts, checking the prices of market items or the status of financial transactions; monitoring traffic conditions or the condition of flights; accessing and updating calendars and schedules; managing reminders, alerts, tasks, and projects; communicating via email or other messaging platforms; and operating devices locally or remotely (eg, making phone calls, controlling lights and temperatures, controlling home security devices, playing music or video, etc.) Enabling the operation of applications and services through natural language dialogue otherwise provided by dedicated applications with graphical user interfaces. In one embodiment, virtual assistant 1002 may be used to launch, operate, and control many of the functions and apps available on the device.

활동들, 제품들, 서비스들, 엔터테인먼트의 소스, 시간 관리, 또는 자연어의 상호작용 대화 및 데이터 및 서비스들에 대한 자동화된 액세스로부터 이익을 얻는 임의의 그외의 유형의 추천 서비스에 대한 개인적 추천들을 제공함.

Providing personal recommendations for activities, products, services, sources of entertainment, time management, or any other type of recommendation service that benefits from interactive conversation in natural language and automated access to data and services .

According to different embodiments, at least some of the various types of functions, operations, actions, and/or other features provided by virtual assistant 1002 may be implemented on one or more client system(s), one or more server(s). may be implemented in the system(s), and/or combinations thereof.

According to different embodiments, at least some of the various types of functions, operations, actions, and/or other features provided by virtual assistant 1002 interpret and actuate user input, as described in greater detail herein. context-related information can be used.

For example, in at least one embodiment, virtual assistant 1002 may be operative to use and/or generate various different types of data and/or other types of information when performing specific tasks and/or actions. have. This may include, for example, input data/information and/or output data/information. For example, in at least one embodiment, virtual assistant 1002 accesses information from one or more different types of sources, devices and/or systems, such as, for example, one or more local and/or remote memory. , process, and/or operate to use. Further, in at least one embodiment, virtual assistant 1002 generates one or more different types of output data/information, which may be stored, for example, in the memory and/or system of one or more local and/or remote devices. can act to make it happen.

Examples of different types of input data/information that may be accessed and/or used by virtual assistant 1002 may include, but are not limited to, one or more (or combinations thereof) of the following.

이동 전화 및 태블릿과 같은 이동 디바이스, 마이크를 구비한 컴퓨터, 블루투스 헤드셋, 자동차 음성 제어 시스템, 전화를 통한 시스템, 응답 서비스 상의 기록, 통합 메시징 서비스 상의 오디오 음성 메일, 클록 라디오와 같은 음성 입력을 갖는 소비자 애플리케이션, 전화국, 홈 엔터테인먼트 제어 시스템, 및 게임 컨솔로부터의 : 음성 입력

Consumers with voice input such as mobile devices such as mobile phones and tablets, computers with microphones, Bluetooth headsets, automotive voice control systems, systems over telephone, record on answering services, audio voice mail on unified messaging services, clock radios Voice input from applications, telephone offices, home entertainment control systems, and game consoles

컴퓨터 또는 이동 디바이스 상의 키보드, 원격 제어 또는 다른 소비자 전자 디바이스 상의 키패드, 비서에 보내지는 이메일 메시지, 비서에 보내지는 인스턴트 메시지 또는 유사한 단문 메시지, 멀티유저 게임 환경에서의 플레이어들로부터 수신한 텍스트, 및 메시지 피드에서 스트리밍되는 텍스트로부터의 텍스트 입력

Keyboards on computers or mobile devices, keypads on remote controls or other consumer electronic devices, email messages sent to assistants, instant messages or similar short messages sent to assistants, texts received from players in a multi-user gaming environment, and messages Text input from text streamed from the feed

센서 또는 위치 기반 시스템으로부터 들어오는 위치 정보. 예시는 글로벌 위치확인 시스템(GPS) 및 이동 전화 상의 어시스트티드 GPS(A-GPS)를 포함한다. 일 실시예에서, 위치 정보는 명확한 사용자 입력과 결합된다. 일 실시예에서, 본 발명의 시스템은 알려진 어드레스 정보 및 현재 위치 결정에 기초하여, 사용자가 집에 있을 때를 검출할 수 있다. 이러한 방식으로, 사용자가 집 밖에 있을 때와는 달리 집에 있을 때에 흥미가 있을 수 있는 정보의 유형뿐만 아니라 사용자가 집에 있는지의 여부에 따라 사용자를 대신해 호출되어야 하는 서비스 및 액션의 유형들에 대한 특정 추론들이 행해질 수 있다.

Location information coming from sensors or location-based systems. Examples include Global Positioning System (GPS) and Assisted GPS on Mobile Phones (A-GPS). In one embodiment, location information is combined with explicit user input. In one embodiment, the system of the present invention is capable of detecting when a user is at home based on known address information and current location determination. In this way, it provides information about the types of information that may be of interest when the user is at home as opposed to when they are away from home, as well as the types of services and actions that should be invoked on behalf of the user depending on whether the user is at home or not. Certain inferences may be made.

클라이언트 디바이스 상의 클록으로부터의 시간 정보. 이는 예를 들면, 로컬 시간 및 시간 존을 가리키는 전화기 또는 다른 클라이언트 디바이스로부터의 시간을 포함할 수 있다. 또한, 시간은 사용자 요청의 내용, 예를 들면 "한 시간 안에" 및 "오늘밤"과 같은 어구를 해석하는 데에 사용될 수 있다.

Time information from the clock on the client device. This may include, for example, time from a phone or other client device indicating the local time and time zone. Also, the time may be used to interpret the content of the user request, for example phrases such as "in an hour" and "tonight".

나침반, 가속도계, 자이로스코프, 및/또는 이동 속도 데이터, 뿐만 아니라 이동 또는 핸드헬드 디바이스로부터의 또는 자동차 제어 시스템과 같은 임베디드 시스템으로부터의 다른 센서 데이터. 이는 원격 제어로부터 응용들 및 게임 콘솔까지의 디바이스 위치확인 데이터를 포함할 수 있다.

Compass, accelerometer, gyroscope, and/or movement speed data, as well as other sensor data from mobile or handheld devices or from embedded systems such as automotive control systems. This may include device location data from remote controls to applications and game consoles.

클릭킹 및 메뉴 선택 및 그래픽 유저 인터페이스(GUI)를 갖는 임의의 디바이스 상의 GUI로부터의 다른 이벤트. 추가 예시들은 터치 스크린에의 터치를 포함한다.

Clicking and menu selections and other events from the GUI on any device with a graphical user interface (GUI). Further examples include touching a touch screen.

센서들 및 알람 클록, 캘린더 알림, 가격 변동 트리거, 위치 트리거, 서버로부터의 디바이스로의 푸시 통지 등과 같은 다른 데이터-구동 트리거로부터의 이벤트.

Events from sensors and other data-driven triggers such as alarm clocks, calendar notifications, price change triggers, location triggers, push notifications from servers to devices, and the like.

Inputs to the embodiments described herein may also include the content of user interaction history, including conversation and request history.

Many different types of output data/information may be generated by virtual assistant 1002 , as described in the relevant cross-referenced US utility model application above. These may include, but are not limited to, one or more of the following (or combinations thereof).

출력 디바이스에 및/또는 디바이스의 사용자 인터페이스에 직접 보내지는 텍스트 출력;

text output directed to the output device and/or to the user interface of the device;

이메일을 통해 사용자에게 보내지는 텍스트 및 그래픽;

text and graphics sent to users via email;

메시징 서비스를 통해 사용자에게 보내지는 텍스트 및 그래픽;

text and graphics sent to users through messaging services;

하기의 하나 이상(또는 그들의 조합)을 포함할 수 있는 음성 출력:

Speech output, which may include one or more (or combinations thereof) of the following:

o synthesized voice;

o sampled speech;

o recorded voice;

사진, 리치 텍스트, 비디오, 사운드, 및 하이퍼링크(예를 들면, 웹 브라우저에 렌더링되는 컨텐츠)를 갖는 정보의 그래픽 레이아웃;

graphic layouts of information with photos, rich text, video, sound, and hyperlinks (eg, content rendered in a web browser);

디바이스가 턴온 또는 턴오프되게 하고, 사운드를 발생시키게 하고, 컬러를 변경하고, 진동하고, 빛을 제어하게 하는 등의 디바이스상의 물리적 작용을 제어하는 액츄에이터 출력;

actuator outputs that control physical actions on the device, such as causing the device to turn on or off, make a sound, change color, vibrate, control light, etc.;

매핑 애플리케이션을 호출, 전화기를 음성 다이얼링, 이메일 또는 인스턴트 메시지를 전송, 미디어를 재생, 캘린더, 태스크 매니저, 및 노트 애플리케이션의 엔트리를 작성, 및 다른 애플리케이션들과 같은 디바이스 상의 다른 애플리케이션을 호출;

invoking mapping applications, voice dialing the phone, sending email or instant messages, playing media, creating entries in calendar, task manager, and notes applications, and calling other applications on the device such as other applications;

원격 카메라, 휠체어의 제어, 원격 스피커 상의 음악 재생, 원격 디스플레이 상의 비디오 재생 등과 같은 디바이스에 부착 또는 제어되는 디바이스들에의 물리적 액션을 제어하는 액츄에이터 출력.

Actuator output that controls physical actions on devices attached or controlled to the device, such as remote camera, control of a wheelchair, playing music on a remote speaker, playing video on a remote display, etc.

It can be appreciated that virtual assistant 1002 of FIG. 27 is one example from the broad range of virtual assistant system embodiments that may be implemented. Other embodiments of a virtual assistant system (not shown) may include, for example, additional, fewer and/or different components/features than those described in the virtual assistant system embodiment of FIG. 27 .

Virtual assistant 1002 may include a plurality of different types of components, devices, modules, processes, systems, etc., which may be implemented and/or illustrated, for example, through the use of hardware and/or combinations of hardware and software. can For example, as shown in the embodiment of FIG. 27 , assistant 1002 may include one or more of the following types of systems, components, devices, processes, etc. (or combinations thereof).

하나 이상의 액티브 온톨로지(ontology)들(1050);

one or more active ontologies 1050;

액티브 입력 유도 컴포넌트(들)(2794)(클라이언트 부분(2794a) 및 서버 부분(2794b)을 포함할 수 있음);

active input induction component(s) 2794 (which may include a client portion 2794a and a server portion 2794b);

단기 개인 메모리 컴포넌트(들)(2752)(마스터 버전(2752b) 및 캐시(2752a)를 포함할 수 있음);

short term personal memory component(s) 2752 (which may include master version 2752b and cache 2752a);

장기 개인 메모리 컴포넌트(들)(2754)(마스터 버전(2754b) 및 캐시(2754a)를 포함할 수 있으며, 예를 들어, 개인 데이터베이스들(1058), 애플리케이션 선호도 및 사용 히스토리(1072) 등을 포함할 수 있음);

Long-term personal memory component(s) 2754 (master version 2754b and cache 2754a), which may include, for example, personal databases 1058, application preferences and usage history 1072, and the like. can);

도메인 모델 컴포넌트(들)(2756);

domain model component(s) 2756;

어휘 컴포넌트(들)(2758)(완전한 어휘(2758b) 및 서브세트(2758a)를 포함할 수 있음);

vocabulary component(s) 2758 (which may include complete vocabulary 2758b and subset 2758a);

언어 패턴 인식기(들) 컴포넌트(들)(2760)(전체 라이브러리(2760b) 및 서브세트(2760a)를 포함할 수 있음);

language pattern recognizer(s) component(s) 2760 (which may include full library 2760b and subset 2760a);

언어 해석기 컴포넌트(들)(2770);

language interpreter component(s) 2770;

도메인 엔티티 데이터베이스(들)(2772);

domain entity database(s) 2772;

대화 플로우 프로세서 컴포넌트(들)(2780);

conversation flow processor component(s) 2780;

서비스 편성 컴포넌트(들)(2782);

service organization component(s) 2782;

서비스 컴포넌트(들)(2784);

service component(s) 2784;

태스크 플로우 모델 컴포넌트(들)(2786);

task flow model component(s) 2786;

대화 플로우 모델 컴포넌트(들)(2787);

dialog flow model component(s) 2787;

서비스 모델 컴포넌트(들)(2788);

service model component(s) 2788;

출력 프로세서 컴포넌트(들)(2790);

output processor component(s) 2790;

In certain client/server based embodiments, some or all of these components may be distributed between the client 1304 and the server 1340 .

In one embodiment, virtual assistant 1002 receives user input 2704 via any suitable input modality, including, for example, touchscreen input, keyboard input, voice input, and/or any combination thereof. do. As described in greater detail herein, in one embodiment, assistant 1002 also receives context information 1000 , which may include event context 2706 and/or any of several other types of context. .

Upon processing user input 2704 and context information 1000 in accordance with the techniques described herein, virtual assistant 1002 generates output 2708 for presentation to the user. Output 2708 may be generated according to any suitable output modality, which may be notified by context 1000 as well as other elements as appropriate. Examples of output modalities are visual output displayed on a screen, auditory output (which may include audio output and/or beeps and other sounds), haptic output (such as vibration), and/or any thereof. includes a combination of

Additional details regarding the operation of the various components shown in FIG. 27 can be found in Related U.S. Patent Application Serial No. 12/987,982, Attorney Docket No. P10575US1, filed Jan. 10, 2011, which is incorporated herein by reference in its entirety. Provided in "Intelligent Automated Assistant".

context

As described above, in one embodiment, virtual assistant 1002 obtains and applies various contextual information to perform information processing functions. The following explanation is given:

가상 비서(1002)에 의해 사용될 컨텍스트 정보의 소스들의 범위

A range of sources of context information to be used by virtual assistant 1002

컨텍스트 정보를 나타내고, 조직하고, 탐색하기 위한 기법들

Techniques for presenting, organizing, and navigating contextual information

컨텍스트 정보가 가상 비서의 몇몇 기능들의 동작을 지원할 수 있는 방법들

How context information can support the operation of some functions of a virtual assistant

분산된 시스템 내에서 컨텍스트 정보를 효과적으로 획득하고, 액세스하고, 적용하기 위한 방법들

Methods for effectively obtaining, accessing, and applying context information within a distributed system

Those skilled in the art will appreciate that the following description of sources, techniques, and methods for using contextual information is illustrative only, and that other sources, techniques and methods may be used without departing from the essential characteristics of the present invention. It will be appreciated that it can be used.

sources of context

Through the steps of information processing performed by virtual assistant 1002 , several different kinds of context may be used to reduce possible interpretations of user input. Examples include application context, personal data context, and previous conversation history. One of ordinary skill in the art will recognize that other sources of context may be available.

Referring now to FIG. 1 , shown is a block diagram illustrating some examples of virtual assistant 1002 , and sources of context that may affect its operation, in accordance with one embodiment. Virtual assistant 1002 takes user input 2704 , such as spoken or typed language, generates output 2708 to the user, and/or performs actions 2710 on behalf of the user. The virtual assistant 1002 shown in FIG. 1 is just one example of a wide variety of virtual assistant system embodiments that may be implemented. Other embodiments (not shown) of the virtual assistant system may include additional, fewer, and/or different components/features than illustrated in the example of virtual assistant 1002 shown in FIG. 1 .

As described in greater detail herein, virtual assistant 1002 may access any of a number of different sources of knowledge and data, such as dictionaries, domain models, and/or task models. In the context of the present invention, such sources, referred to as background sources, are internal to assistant 1002 . In addition to user input 2704 and background sources, virtual assistant 1002 may also include, for example, device sensor data 1056 , application preferences and usage history 1072 , conversation history and secondary memory 1052 , personal database 1058 . ), personal acoustic context data 1080 , current application context 1060 , and event context 2706 . This will be described in detail herein.

Application Context (1060)

Application context 1060 refers to the state of an application or similar software in which the user is doing something. For example, a user may use a text messaging application to chat with a specific person. Virtual assistant 1002 need not be specific to or be part of a user interface of a text messaging application. Rather, virtual assistant 1002 may receive context from any number of applications, and each application may pass its context to inform virtual assistant 1002 .

If the user is currently using an application when virtual assistant 1002 is invoked, the state of that application can provide useful contextual information. For example, when virtual assistant 1002 is invoked from within an email application, context information may include sender information, recipient information, sent date and/or time, subject, data extracted from the email content, mailbox or folder name, etc. may include

11-13 , shown is a set of screen shots illustrating an example of deriving what a pronoun refers to using an application context in a text messaging domain, according to one embodiment. 11 shows a screen 1150 that may be displayed while a user is using a test messaging application. 12 shows screen 1250 after virtual assistant 1002 is activated in the context of a text messaging application. In this example, virtual assistant 1002 presents prompt 1251 to the user. In one embodiment, the user may provide spoken input by tapping the microphone icon 1252 . In another embodiment, assistant 1002 may accept voice input at any time and does not require the user to tap microphone icon 1252 before providing input; Accordingly, icon 1252 may be a reminder that assistant 1002 is waiting for voice input.

In FIG. 13 , as shown on screen 1253 , the user is having a conversation with virtual assistant 1002 . The input "call him" spoken by the user has been echoed back, and virtual assistant 1002 is responding that it will call a specific person at a specific phone number. To interpret the user's ambiguous input, virtual assistant 1002 combines and uses contexts from various sources to derive what the pronoun refers to, as described in more detail herein.

17-20 , another example of enabling a command to be interpreted and manipulated using a current application context is illustrated, according to an embodiment.

In FIG. 17 , the user is presented with the user's email inbox 1750 , and the user selects to view a particular email message 1751 . 18 shows an email message 1751 after an email message to be viewed has been selected; In this example, email message 1751 includes an image.

In FIG. 19 , the user activated virtual assistant 1002 while viewing an email message 1751 from within an email application. In one embodiment, the email message 1751 is moved above the screen to be displayed, making room for the prompt 150 from the virtual assistant 1002 . This display reinforces the notion that virtual assistant 1002 is providing assistance in the context of the currently viewed email message 1751 . Accordingly, the user's input to virtual assistant 1002 will be interpreted in the current context of the email message 1751 being viewed.

In Figure 20, the user has provided a command 2050: "Reply let's get this to marketing right away". Contextual information is used to interpret command 2050 , including email message 1751 and information about the email application in which the email message is displayed. This context can be used to determine the meaning of the words "reply" and "this" in command 2050 , and how to set up an email configuration transaction for a particular recipient on a particular message thread. In this case, virtual assistant 1002 can access the context information and determine that "marketing" refers to a recipient named "John Applecore" and can determine an email address to use for the recipient. Thus, virtual assistant 1002 composes email 2052 that the user approves and sends. In this way, virtual assistant 1002 may enable manipulation of tasks (constructing an email message) based on user input along with contextual information describing the state of the current application.

Application context can also help identify the meaning of user intent across applications. Referring to FIG. 21 , in the context of an email being viewed (eg, email message 1751 ), the user calls virtual assistant 1002 , and the user's command 2150 says "Send him a text..." An example is shown. Command 2150 is interpreted as indicating that a text message rather than an email should be sent by virtual assistant 1002 . However, the use of the word "him" indicates that the same recipient (John Appleseed) is intended. Thus, virtual assistant 1002 recognizes that this communication should go to a different channel (text message to the person's phone number obtained from contact information stored on the device) to this recipient. Accordingly, the virtual assistant 1002 composes a text message 2152 that the user approves and transmits.

Examples of context information that may be obtained from application(s) include, but are not limited to:

애플리케이션의 아이덴티티;

the identity of the application;

현재 이메일 메시지, 현재 음악 또는 재생목록 또는 재생되는 채널, 현재 책 또는 영화 또는 사진, 현재 캘린더 일/주/달, 현재 리마인더 리스트, 현재 전화 통화, 현재 텍스트 메시징 대화, 현재 지도 위치, 현재 웹 페이지 또는 검색 쿼리, 위치 감지 애플리케이션들을 위한 현재 도시 또는 다른 위치, 현재 소셜 네트워크 프로필, 또는 현재 객체들의 임의의 다른 애플리케이션 특정 개념;

Current email message, current music or playlist or playing channel, current book or movie or photo, current calendar day/week/month, current reminder list, current phone call, current text messaging conversation, current map location, current web page or a search query, current city or other location for location sensing applications, current social network profile, or any other application specific concept of current objects;

이름, 장소, 날짜, 및 현재 객체들로부터 추출될 수 있는 식별가능한 다른 엔티티 또는 값.

Name, place, date, and other identifiable entity or value that can be extracted from the current objects.

Personal databases (1058)

Another source of context data is the user's personal database(s) 1058 on a device such as a phone, such as an address book containing names and phone numbers. Referring to FIG. 14 , there is shown an example of a screen shot 1451 in which virtual assistant 1002 prompts for an ambiguous name, in accordance with one embodiment. Here, the user said "Call Herb"; Virtual assistant 1002 prompts the user to select from among matching contacts in the user's address book. Thus, the address book is used as a source of personal data context.

In one embodiment, the user's personal information is obtained from the personal database 1058 for use as a context for enabling interpretation and/or manipulation of the user's intentions or other functions of the virtual assistant 1002 . For example, data in a user's contacts database can be used to reduce ambiguity in interpreting a user's commands when the user mentions someone by name only. Examples of contextual information that may be obtained from personal databases 1058 include, but are not limited to:

사용자의 연락처 데이터베이스(어드레스 북) -- 이름, 전화 번호, 물리 어드레스, 네트워크 어드레스, 계정 식별자, 중요한 날짜에 관한 정보가 포함됨 -- 사람, 회사, 조직, 위치, 웹 사이트, 및 사용자가 참조할 수 있는 다른 엔티티에 대한 것;

A user's contact database (address book) -- contains information about names, phone numbers, physical addresses, network addresses, account identifiers, and important dates -- people, companies, organizations, locations, websites, and can be referenced by users for other entities in the;

사용자 자신의 이름, 선호되는 발음, 어드레스, 전화 번호, 등;

the user's own name, preferred pronunciation, address, phone number, etc.;

어머니, 아버지, 자매, 상사 등과 같은 사용자가 명명한 관계;

User-named relationships such as mother, father, sister, boss, etc.;

캘린더 이벤트, 특별한 날의 명칭, 사용자가 참조할 수 있는 명명된 다른 엔트리들을 포함하는, 사용자의 캘린더 데이터;

the user's calendar data, including calendar events, names of special days, and other named entries to which the user may refer;

해야 하거나 기억해야 하거나 얻어야 할 것의 목록을 포함하여 사용자가 참조할 수 있는 사용자의 리마인더 또는 작업 목록;

a list of reminders or tasks for the user to which the user may refer, including a list of things to do, remember, or obtain;

노래, 장르, 재생 목록, 및 사용자가 참조할 수 있는 사용자의 뮤직 라이브러리와 연관된 다른 데이터;

songs, genres, playlists, and other data associated with the user's music library to which the user may refer;

사람, 장소, 카테고리, 태그, 레이블, 또는 사용자의 미디어 라이브러리 내의 포토 또는 비디오 또는 다른 미디어 상의 다른 상징적 이름;

a person, place, category, tag, label, or other symbolic name on a photo or video or other media within the user's media library;

제목, 저자, 장르, 또는 사용자의 개인 라이브러리에 있는 다른 문헌 또는 어드레스 북의 다른 상징적 이름 .

Title, author, genre, or other symbolic name of another document in the user's personal library or address book.

Conversation History (1052)

Another source of context data is the conversation history 1052 of the virtual assistant 1002 and the user. Such history may include, for example, references to domains, people, places, and the like. Referring to FIG. 15 , an example is shown in which virtual assistant 1002 infers a location for a command using a conversation context, according to one embodiment. At screen 1551, the user first asks "What's the time in New York"; Virtual assistant 1002 responds 1552 by providing the current time in New York City. Afterwards the user asks "What's the weather". The virtual assistant 1002 uses the previous conversation history to infer that the location the weather query intended is the last location mentioned in the conversation history. Accordingly, the response 1553 provides weather information for New York City.

As another example, if the user says "find camera shops near here", after examining the results, says "how about in San Francisco?", and the assistant uses the conversation context to "do the The same task (find camera stores)" is determined, and the meaning of "in San Francisco" is determined to be "changing the locus of the search from here to San Francisco." Virtual assistant 1002 may also use as context the details of a previous conversation, such as previous output provided to the user. For example, if virtual assistant 1002 used a witty response intended as humor, such as "Sure thing, you're the boss," the virtual assistant remembers that it has already said such a response and, within the conversation session, You can avoid repeating such phrases.

Examples of context information from conversation history and virtual auxiliary memory include:

대화에서 언급된 사람들;

those mentioned in the conversation;

대화에서 언급된 장소들 및 위치들;

places and locations mentioned in the conversation;

포커스되는 현재의 시간 프레임;

the current time frame being focused;

이메일이나 캘린더와 같이, 포커스되는 현재의 애플리케이션 도메인;

the current application domain in focus, such as email or calendar;

이메일을 읽는 것 또는 캘린더 엔트리를 생성하는 것과 같이, 포커스되는 현재의 태스크;

the current task being focused on, such as reading an email or creating a calendar entry;

막 읽은 이메일 메시지 또는 막 생성한 캘린더 엔트리와 같이, 포커스되는 현재의 도메인 객체들;

current domain objects in focus, such as an email message you just read or a calendar entry you just created;

질문을 하고 있는지 및 어떤 가능한 대답들이 예상되는지와 같이, 대화 또는 트랜잭션 플로우의 현재의 상태;

the current state of the conversation or transaction flow, such as whether a question is being asked and what possible answers are expected;

"good Italian restaurants"와 같이, 사용자 요청들의 히스토리;

history of user requests, such as "good Italian restaurants";

리턴된 레스토랑 세트와 같이, 사용자 요청들의 결과들의 히스토리;

a history of the results of user requests, such as a returned restaurant set;

대화에서 비서에 의해 이용된 어구들의 히스토리;

a history of phrases used by the secretary in the conversation;

"my mother is Rebecca Richards" 및 "I liked that restaurant"와 같이, 사용자가 비서에게 말한 사실들

Facts the user told the secretary, such as "my mother is Rebecca Richards" and "I liked that restaurant"

includes, but is not limited to.

Referring now to FIGS. 25A and 25B , shown is a series of screenshots illustrating an example of use of a previous conversation context in accordance with one embodiment. In FIG. 25A , the user has entered a request 2550 for any new email from John. Virtual assistant 1002 responds by displaying email message 2551 from John. In FIG. 25B , the user enters command 2552 "Reply let's get this to marketing right away". Virtual assistant 1002 interprets command 2552 using the previous conversation context; Specifically, this command is interpreted to refer to the email message 2551 displayed in FIG. 25 .

Device Sensor Data (1056)

In one embodiment, a physical device executing virtual assistant 1002 may have one or more sensors. These sensors may provide sources of contextual information. Examples of such information include:

사용자의 현재 위치;

your current location;

사용자의 현재 위치에서의 현지 시간;

local time at the user's current location;

디바이스의 위치, 방위 및 모션;

the position, orientation and motion of the device;

현재의 광 레벨; 온도 및 다른 환경 척도;

current light level; temperature and other environmental measures;

사용 중인 마이크로폰들 및 카메라들의 속성들;

properties of the microphones and cameras in use;

이용되고 있는 현재의 네트워크들 및 접속된 네트워크들의 서명들 - 이 네트워크는 이더넷, Wi-Fi 및 블루투스를 포함함 -

Signatures of current networks being used and connected networks - this network includes Ethernet, Wi-Fi and Bluetooth -

includes, but is not limited to. Signatures include MAC addresses of network access points, assigned IP addresses, device identifiers such as Bluetooth names, frequency channels and other attributes of wireless networks.

The sensors may be any type of sensor including, for example, an accelerometer, compass, GPS unit, altitude detector, light sensor, thermometer, barometer, clock, network interface, battery test circuit, and the like.

Application Preference and Usage History (1072)

In one embodiment, information describing the user's usage history 1072 as well as the user's preferences and settings for various applications interprets and/or manipulates the user's intent or other function of the virtual assistant 1002 . It is used as a context to enable Examples of such preferences and history 1072 include:

단축키, 즐겨찾기, 북마크, 친구 리스트, 또는 사람들, 회사들, 어드레스들, 전화 번호들, 장소들, 웹 사이트들, 이메일 메시지들 또는 임의의 다른 참조들에 대한 사용자 데이터의 임의의 다른 집합;

shortcuts, favorites, bookmarks, friends list, or any other collection of user data for people, companies, addresses, phone numbers, places, web sites, email messages or any other references;

디바이스 상에서 이루어진 최근의 통화들;

recent calls made on the device;

대화 상대방들을 포함하는 최근의 텍스트 메시지 대화들;

recent text message conversations involving chat parties;

지도 또는 방향에 대한 최근의 요청들;

recent requests for maps or directions;

최근의 웹 검색들 및 URL들;

recent web searches and URLs;

주식 애플리케이션에 열거된 주식들;

stocks listed in the stock application;

재생되는 최근의 노래나 비디오 또는 다른 미디어;

the most recent song or video or other media being played;

경보 애플리케이션들 상에 설정된 알람의 명칭들;

names of alarms set on alarm applications;

디바이스 상의 애플리케이션들 또는 다른 디지털 객체들의 명칭들;

names of applications or other digital objects on the device;

사용자의 선호 언어 또는 사용자의 위치에서 사용 중인 언어

Your preferred language or the language your location is using

includes, but is not limited to.

Referring now to Figure 16, an example of the use of a phone favorites list as a source of context is shown in accordance with one embodiment. On screen 1650 , a list of favorite contacts 1651 is shown. When the user provides input for "call John", a list of these favorite contacts 1651 can be used to determine that "John" refers to John Appleseed's mobile number because that number is Because it appears on this list.

Event Context (2706)

In one embodiment, virtual assistant 1002 may utilize context associated with asynchronous events that occur independently of the user's interaction with virtual assistant 1002 . 22 to 24 , an example illustrating activation of the virtual assistant 1002 after an event that may provide an event context or an alert context according to an embodiment occurs is illustrated. In this case, the event is an incoming text message 2250 as shown in FIG. In FIG. 23 , virtual assistant 1002 has been invoked, and text message 2250 appears along with prompt 1251 . In FIG. 24 , the user has entered the command “call him” 2450 . Virtual assistant 1002 uses the event context to disambiguate the command by interpreting "him" to mean the person who sent the incoming text message 2250 . Virtual assistant 1002 also uses the event context to determine which phone number to use for outgoing calls. A confirmation message 2451 is displayed to indicate that a call is being made.

Examples of alert context information include:

착신 텍스트 메시지들 또는 페이지들;

incoming text messages or pages;

착신 이메일 메시지들;

incoming email messages;

착신 전화 통화들;

incoming phone calls;

리마인더 통지들 또는 태스크 경보들;

reminder notifications or task alerts;

캘린더 경보들;

calendar alerts;

알람 클록, 타이머 또는 다른 시간-기반 경보들;

alarm clock, timer or other time-based alarms;

게임으로부터의 스코어나 다른 이벤트의 통지들;

notifications of scores or other events from the game;

주가 경보와 같은 금융 이벤트의 통지들;

notifications of financial events, such as stock price alerts;

뉴스 플래시들 또는 다른 방송 통지들;

news flashes or other broadcast notifications;

임의의 애플리케이션으로부터의 푸시 통지들

Push notifications from any application

includes, but is not limited to.

Personal Acoustic Context Data (1080)

When interpreting the voice input, the virtual assistant 1002 may also consider the acoustic environment in which the voice is input. For example, the noise profile of a quiet office is different from that of a car or public place. If the speech recognition system is capable of identifying and storing acoustic profile data, such data may also be provided as contextual information. When combined with other contextual information such as the properties of the microphone being used, current location and current conversational state, acoustic context can support recognition and interpretation of input.

Representation and access to context

As noted above, virtual assistant 1002 may utilize contextual information from any of a number of different sources. Any of a number of different mechanisms may be used to represent context to be made available to virtual assistant 1002 . Referring now to Figures 8A-8D, there are shown several examples of representations of context information as may be used in connection with various embodiments of the present invention.

Representation of people, places, times, domains, tasks and objects

8A shows examples 801-809 of context variables representing simple attributes, such as the geographic coordinates of a user's current location. In one embodiment, a current value may be maintained for a core set of context variables. For example, there may be a current user, a current location of focus, a current time frame that is focused, a current application domain that is focused, a current task that is focused, and a current domain object that is focused. A data structure as shown in Figure 8a may be used for this representation.

8B shows an example 850 of a more complex representation that may be used to store contextual information for a contact. Also shown is an example 851 of an expression containing data for a contact. In one embodiment, a contact (or person) may be represented as an object with attributes for name, gender, address, phone number, and other attributes that may be maintained in a contact database. Similar representations may be used for place, time, application domain, task, domain object, and the like.

In one embodiment, a set of current values of a given type is represented. Such a set may refer to current people, current places, current times, and the like.

In one embodiment, context values are arranged in history so that, at iteration N, there is a frame of current context values, and there is also a frame of context values that were current at iteration N-1, which is the required length of history. certain restrictions are placed on it. 8C shows an example of an array 811 containing a history of context values. Specifically, each column in FIG. 8C represents a context variable, where the rows correspond to different times.

In one embodiment, the set of typed context variables is arranged in history as shown in FIG. 8D . In this example, a set of context variables 861 referring to people is shown along with another set 871 of context variables referring to places. Accordingly, relevant context data for a specific time in history can be retrieved and applied.

Those skilled in the art will recognize that the specific representations shown in FIGS. 8A-8D are exemplary only, and that many other mechanisms and/or data formats for representing context may be used. Examples include:

일 실시예에서는, 가상 비서(1002)가 사용자에게 어떻게 어드레스하고 사용자의 집, 업무, 이동 전화 등을 참조할지를 알 수 있도록 시스템의 현재 사용자가 일부 특수한 방식으로 표현될 수 있다.

In one embodiment, the current user of the system may be represented in some special way so that virtual assistant 1002 knows how to address the user and refer to the user's home, work, mobile phone, etc.

일 실시예에서는, 가상 비서(1002)가 "나의 엄마" 또는 "나의 상사의 집"과 같은 참조(reference)들을 이해할 수 있도록 사람들 간의 관계가 표현될 수 있다.

In one embodiment, relationships between people may be represented so that virtual assistant 1002 can understand references such as "my mom" or "my boss's house".

장소는, 이름, 거리 어드레스, 지리적 좌표 등과 같은 속성들을 갖는 객체로서 표현될 수 있다.

A place may be represented as an object having properties such as name, street address, geographic coordinates, and the like.

시간은 (년, 월, 일, 시간, 분 또는 초 등의) 유니버셜 타임, 타임존 오프셋, 레졸루션을 포함하는 속성들을 갖는 객체로서 표현될 수 있다. 시간 객체는 또한 "오늘", "이번 주", "이번 (다가오는) 주말", "다음 주", "애니의 생일" 등과 같은 심볼 시간을 나타낼 수도 있다. 또한, 시간 객체는 시기들 또는 시점들을 나타낼 수 있다.

Time may be expressed as an object having properties including universal time (such as year, month, day, hour, minute or second), time zone offset, and resolution. Time objects may also represent symbol times such as "today", "this week", "this (upcoming) weekend", "next week", "Annie's birthday", and the like. Also, a time object may represent times or points in time.

이메일, 텍스트 메시징, 전화, 캘린더, 연락처, 사진, 비디오, 지도, 날씨, 리마인더, 시계, 웹 브라우저, 페이스북, 판도라 등과 같은 담화(discourse)의 도메인 또는 애플리케이션 또는 서비스를 나타내는 애플리케이션 도메인의 면에서 또한 컨텍스트가 제공될 수 있다. 현재의 도메인은 어떤 도메인에 초점이 맞추어져 있는지를 표시한다.

Also in terms of domains of discourse such as email, text messaging, phone, calendar, contacts, photos, videos, maps, weather, reminders, clocks, web browsers, Facebook, Pandora, etc. or application domains representing applications or services. Context may be provided. The current domain indicates which domain is in focus.

컨텍스트는 또한 도메인 내에서 수행하기 위한 하나 이상의 태스크, 또는 동작들을 정의할 수 있다. 예를 들어, 이메일 도메인 내에서는, 이메일 메시지 판독, 이메일 검색, 새로운 이메일 작성 등과 같은 태스크들이 있다.

A context may also define one or more tasks, or actions, to perform within a domain. For example, within an email domain, there are tasks such as reading email messages, retrieving emails, composing new emails, and the like.

도메인 객체들은 다양한 도메인과 연관된 데이터 객체들이다. 예를 들면, 이메일 도메인은 이메일 메시지 상에서 동작하며, 캘린더 도메인은 캘린더 이벤트 상에서 동작한다.

Domain objects are data objects associated with various domains. For example, an email domain operates on email messages, and a calendar domain operates on calendar events.

For the purposes of the description provided herein, a representation of context information is referred to as a context variable of a given type. For example, the representation of the current user is a context variable of type person.

Representation of context derivation

In one embodiment, the derivation of the context variable is explicitly expressed so that it can be used for information processing. The derivation of contextual information characterizes the source and/or sets of interferences from which the information is concluded and retrieved. For example, the person context value 851 shown in FIG. 8B may be derived from a text message domain object obtained from the event context 2706 . The source of this context value 851 may be represented.

Representation of the history of user requests and/or intents

In one embodiment, a history of user requests may be stored. In one embodiment, a history of profoundly structured representations of user intent (derived from natural language processing) may also be stored. This allows virtual assistant 1002 to make new input meaningful in the context of previously interpreted input. For example, if the user asks "what's the weather like in New York?", language interpreter 2770 may interpret the question as referring to a location in New York. Then, when the user says "How is this week?", the virtual assistant 1002 may refer to the previous interpretation and determine that "how is" should be interpreted as meaning "how is the weather".

Representation of result history

In one embodiment, the history of the results of the user request may be stored in the form of a domain object. For example, for a user's request, "Find a good Italian restaurant," a set of domain objects representing the restaurant may be returned. Thereafter, when the user inputs a command such as "Call Amilios," the virtual assistant 1002, in the search result, displays a restaurant named Amilios, which is the smallest set of all possible places that can be called up. You can search results for .

Deferred binding of context variables

In one embodiment, a context variable may represent information that is retrieved or derived on demand. For example, a context variable representing the current location, when accessed, may call an API that retrieves the current location data from the device, and then perform other processing to, for example, calculate a street address. The value of that context variable may be maintained for a specific period of time, depending on the caching policy.

Context Search

Virtual assistant 1002 may use any of many different approaches for retrieving relevant contextual information to solve information processing problems. Examples of different types of searches include, but are not limited to:

컨텍스트 변수 이름에 의한 검색. "현재 사용자의 이름(first name)"과 같이, 요구된 컨텍스트 변수의 이름이 알려져 있는 경우, 가상 비서(1002)는 그 이름의 인스턴스를 검색할 수 있다. 히스토리가 유지되고 있는 경우, 가상 비서(1002)는 먼저 현재 값들을 검색한 다음, 매칭이 발견될 때까지 이전 데이터를 컨설팅한다.

Search by context variable name . If the name of the requested context variable is known, such as "current user's first name," virtual assistant 1002 can retrieve an instance of that name. If a history is being maintained, virtual assistant 1002 first retrieves the current values and then consults the previous data until a match is found.

컨텍스트 변수 유형에 의한 검색. 개인과 같은, 요청된 컨텍스트 변수의 유형이 알려져 있는 경우, 가상 비서(1002)는 이러한 유형의 컨텍스트 변수의 인스턴스들을 검색할 수 있다. 히스토리가 유지되는 경우, 가상 비서(1002)는 먼저 현재 값들을 검색한 다음, 매칭이 발견될 때까지 이전 데이터를 컨설팅할 수 있다.

Search by context variable type . If the type of the requested context variable is known, such as an individual, virtual assistant 1002 can retrieve instances of this type of context variable. If a history is maintained, virtual assistant 1002 may first retrieve the current values and then consult the previous data until a match is found.

In one embodiment, if the current information processing problem requires only one match, the search ends once a match is found. If multiple matches are allowed, matching results may be searched until some limit is reached.

In one embodiment, if appropriate, virtual assistant 1002 may limit searches to data with specific derivations. For example, when looking for a person object within a task flow to send an email, virtual assistant 1002 may only need to consider a context variable whose derivation is an application associated with that domain.

In one embodiment, virtual assistant 1002 uses rules to rank matches according to heuristics using any available properties of context variables. For example, when processing a user input including a command "Tell her I'm late", the virtual assistant 1002 interprets "her" with reference to the context. In the meantime, virtual assistant 1002 may apply sequencing to display preferences for human objects whose derivation is application usage history for communication applications such as text messaging and email. As another example, when interpreting the command "Call her", the virtual assistant 1002 may apply the ranking as favoring Person objects with phone numbers over Person objects with unknown phone numbers. have. In one embodiment, sequencing rules may be associated with domains. For example, different sequencing rules may be used for ranking personal variables for email and phone domains. Those skilled in the art will recognize that any such sequencing rule(s) may be created and/or applied depending on the particular representation and access to the contextual information needed.

Use of context to enhance virtual assistant processing

As described above, context may be applied to various calculations and inferences in relation to the operation of virtual assistant 1002 . Referring now to FIG. 2 , a flow diagram is shown illustrating a method 10 of using context in various processing steps in virtual assistant 1002 , according to one embodiment.

Method 10 may be implemented in connection with one or more embodiments of virtual assistant 1002 .

In at least one embodiment, method 10 may be operated to perform various types of functions, operations, actions, and/or other features, such as, for example, one or more steps (or combinations thereof) below. can:

사용자와 가상 비서(1002) 간의 대화 인터페이스의 인터페이스 제어 흐름 루프를 실행한다. 방법(10)의 적어도 한 번의 반복은 대화에서 플라이(ply)로서 역할을 할 수 있다. 대화 인터페이스는, 대화 방식으로, 발성을 앞뒤로 함으로써 사용자와 비서(1002)가 통신하는 인터페이스이다.

Executes the interface control flow loop of the dialog interface between the user and the virtual assistant 1002 . At least one iteration of method 10 may serve as a ply in the conversation. The conversational interface is an interface through which the user and the assistant 1002 communicate by uttering back and forth in a conversational manner.

가상 비서(1002)에 대한 실행적인 제어 흐름을 제공한다. 즉, 절차는 입력의 수집, 입력의 처리, 출력의 생성, 및 출력의 사용자에의 제시를 제어한다.

Provides an executable control flow for virtual assistant 1002 . That is, the procedure controls the collection of input, processing of the input, generation of the output, and presentation of the output to the user.

가상 비서(1002)의 컴포넌트들 간의 통신을 조정한다. 즉, 하나의 컴포넌트의 출력이 다른 하나의 컴포넌트에 피딩되는 장소 그리고 환경으로부터의 전체 입력 및 환경에 대한 액션이 발생할 수 있는 장소를 지시할 수 있다.

Coordinates communications between components of virtual assistant 1002 . That is, it is possible to indicate a place where the output of one component is fed to the other component, and a place where the entire input from the environment and an action on the environment can occur.

In at least some embodiments, portions of method 10 may also be implemented in other devices and/or systems of a computer network.

According to a particular embodiment, multiple instances or threads of method 10 may be implemented concurrently and/or initiated concurrently by use of one or more processors 63 and/or other combinations of hardware and/or hardware and software. have. In at least one embodiment, one or more or selected portions of method 10 may be implemented in one or more client(s) 1304 , one or more server(s) 1340 , and/or combinations thereof.

For example, in at least some embodiments, various aspects, features, and/or functionalities of method 10 may be performed, implemented and / or may be initiated.

According to different embodiments, one or more threads or instances of method 10 may be of one or more different types (eg, such as a minimum threshold criterion) to trigger initiation of at least one instance of method 10 . may be initiated in response to detection of one or more conditions or events that satisfy a criterion. Examples of various types of conditions or events that may trigger the initiation and/or implementation of examples of one or more different threads or instances of a method may include, but are not limited to, one or more (or combinations thereof) of the following: :

예를 들어, 다음 중 하나 이상(이에 제한되지 않음)과 같은 가상 비서(1002)의 인스턴스를 갖는 사용자 세션:

For example, a user session with an instance of virtual assistant 1002 such as but not limited to one or more of the following:

o a mobile device application starting up, for example, a mobile device application implementing an embodiment of virtual assistant 1002;

o a computer application launching, for example, an application implementing an embodiment of virtual assistant 1002;

o a dedicated button on the mobile device to be pressed, such as a “voice input button”;

o a button on a peripheral device attached to a computer or mobile device, such as a headset, telephone handset or base station, GPS navigation system, consumer appliance, remote control or any other device having a button that may be associated with calling an assistant;

o a web session initiated from a web browser to a website implementing the virtual assistant 1002;

o an interaction initiated from within an existing web browser session, for example, with a website implementing virtual assistant 1002 for which virtual assistant 1002 service is requested;

o an email message sent to the modality server 1426 that coordinates communications with the implementation of the virtual assistant 1002;

o A text message is sent to the modality server 1426 that coordinates communication with the implementation of the virtual assistant 1002;

o a phone call is made to the modality server 1434 that coordinates communication with the implementation of the virtual assistant 1002;

o An event, such as an alert or notification, is sent to the application providing the implementation of the virtual assistant 1002.

가상 비서(1002)를 제공하는 디바이스가 턴온 및/또는 시작될 때.

When the device providing the virtual assistant 1002 is turned on and/or started.

According to different embodiments, one or more different threads or instances of method 10 may be initiated and/or implemented manually, automatically, statistically, dynamically, concurrently and/or combinations thereof. Further, different instances and/or embodiments of method 10 may be initiated in one or more different time intervals (eg, during specific time intervals, at regular periodic intervals, at irregular periodic intervals, on demand). etc).

In at least one embodiment, a given instance of method 10 uses and/or uses a variety of different types of data and/or other types of information when performing certain tasks and/or actions, including contextual data as described above. Or you can create The data may also include any other types of input data/information and/or output data/information. For example, in at least one embodiment, at least one instance of method 10 may access, process, and/or otherwise use information from one or more different types of sources, such as, for example, one or more databases. . In at least one embodiment, at least a portion of the database information may be accessed via communication with one or more local and/or remote memory devices. Further, at least one instance of method 10 may generate one or more different types of output data/information, which may be stored in, for example, local memory and/or remote memory devices.

In at least one embodiment, initial configuration of a given instance of method 10 may be performed using one or more different types of initialization parameters. In at least one embodiment, at least some of the initialization parameters may be accessed via communication with one or more local and/or remote memory devices. In at least one embodiment, at least some of the initialization parameters provided to an instance of method 10 may correspond to, and/or be derived from, input data/information.

In the specific example of FIG. 2 , assume a single user is accessing an instance of virtual assistant 1002 over a network from a client application with voice input capabilities.

The voice input is elicited and interpreted (100). Induction may include presenting prompts in any suitable mode. In various embodiments, the client's user interface provides several modes of input. These may include, for example;

액티브 타이핑된 입력(active typed-input) 유도 절차를 호출할 수 있는, 타이핑된 입력을 위한 인터페이스;

an interface for typed input, capable of invoking an active typed-input derivation procedure;

액티브 음성 입력 유도 절차를 호출할 수 있는, 음성 입력을 위한 인터페이스;

an interface for voice input capable of invoking an active voice input elicitation procedure;

액티브 GUI 기반 입력 유도를 호출할 수 있는, 메뉴로부터 입력들을 선택하기 위한 인터페이스.

An interface for selecting inputs from a menu, capable of invoking an active GUI based input elicitation.

Techniques for performing each of these are disclosed in the relevant patent applications referenced above. Those skilled in the art will recognize that other input modes may be provided. The output of step 100 is a set 190 of candidate interpretations of the input speech.

A set 190 of candidate interpretations is processed 200 by a language interpreter 2770 (also called a natural language processor or NLP), which parses the text input and a set 290 of possible interpretations of user intent. ) is created.

At step 300 , the indication(s) 290 of the user's intent is communicated to a dialog flow processor 2780 implementing an embodiment of the dialog and flow analysis procedure described with respect to FIG. 5 . The dialog flow processor 2780 determines which interpretation of intent is most likely, maps this interpretation to instances of domain models and parameters of the task model, and determines the next flow step in the task flow.

At step 400, the identified flow step is executed. In one embodiment, the invocation of the flow step is performed by the service orchestration component 2782 which invokes the set of services on behalf of the user's request. In one embodiment, these services contribute some data to a common result.

At step 500, a dialog response is generated. In step 700, a response is sent to the client device for output. The client software on the device renders this on the screen (or other output device) of the client device.

After viewing the response, if the user completes 790, the method ends. If the user does not exit, another iteration of the loop is initiated by returning to step 100 .

Context information 1000 may be used by various components of the system at various points in method 10 . For example, as shown in FIG. 2 , context 1000 may be used in steps 100 , 200 , 300 and 500 . A further description of the use of context 1000 in these steps is provided below. However, those skilled in the art will recognize that the use of the context information is not limited to these specific steps, and that the system may use the context information at other points without departing from the essential features of the present invention.

In addition, those skilled in the art will appreciate that different embodiments of method 10 may include additional features and/or operations other than those shown in the specific embodiment shown in FIG. It will be appreciated that at least some of the features and/or operations of the same method 10 may be omitted.

Use of Context in Speech Guidance and Interpretation

Referring now to FIG. 3 , shown is a flow diagram illustrating a method for using context in speech derivation and interpretation 100 to enhance speech recognition in accordance with one embodiment. Context 1000 may be used for disambiguation in speech recognition, for example, to guide generation, ranking, and filtering of candidate hypotheses that match phonemes to words. Different speech recognition systems use various mixtures of generation, sequence, and filter, but context 1000 can generally be applied to reduce hypothesis space at any stage.

The method begins (100). The virtual assistant 1002 receives a voice or voice input in the form of an auditory signal ( 121 ). The speech-to-text service 122 or processor generates a set of candidate text interpretations 124 of the auditory signal. In one embodiment, the speech-to-text conversion service 122 is implemented using, for example, Nuance Recognizer available from Nuance Communications, Inc. of Burlington, Massachusetts.

In one embodiment, virtual assistant 1002 uses statistical language models 1029 to generate candidate text interpretation 124 of speech input 121 . In one embodiment, context 1000 is applied to bias the generation, filtering, and/or sequencing of candidate interpretations 124 generated by speech-to-text conversion service 122 . E.g,

음성-텍스트 변환 서비스(122)는 통계 언어 모델들(1029)을 바이어스하기 위하여 사용자 퍼스널 데이터베이스(들)(1058)로부터의 어휘를 사용할 수 있다.

Speech-to-text service 122 may use vocabulary from user personal database(s) 1058 to bias statistical language models 1029 .

음성-텍스트 변환 서비스(122)는 커스텀 통계 언어 모델(1029)을 선택하기 위하여 대화 상태 컨텍스트를 사용할 수 있다. 예를 들어, 예/아니오 질문을 할 때, 이들 단어들을 듣는 쪽으로 바이어스되는 통계 언어 모델(1029)이 선택될 수 있다.

The speech-to-text service 122 may use the conversation state context to select a custom statistical language model 1029 . For example, when asking a yes/no question, a statistical language model 1029 may be selected that is biased toward hearing these words.

음성-텍스트 변환 서비스(122)는 적절한 단어들 쪽으로 바이어스하기 위하여 현재 애플리케이션 컨텍스트를 사용할 수 있다. 예를 들어, "그녀에게 전화하는 것(call her)"는 텍스트 메시지 애플리케이션 컨텍스트에서 "칼라(collar)"보다 선호(prefer)될 수 있는데, 왜냐하면 이러한 컨텍스트는 전화걸 수 있는 사람 객체(Person Object)들을 제공하기 때문이다.

Speech-to-text service 122 may use the current application context to bias towards appropriate words. For example, "call her" may be preferred over "collar" in the context of a text message application, because this context is a Person Object that can be called because they provide

For example, a given speech input may elicit the speech-to-text conversion service 122 to generate interpretations "call her" and "color". A speech-to-text service 122, guided by Statistical Language Models (SLM) 1029, may be tuned by syntactic constraints to hear names after hearing "call". can Speech-to-text service 122 may also be tuned based on context 1000 . For example, if "Herb" is the name of a user address book, this context can be used to lower the threshold to consider "Herb" as an interpretation of the second syllable. That is, the presence of names in the context of a user's personal data may affect the selection and tuning of the statistical language model 1029 used to build hypotheses. The name “Herb” may be part of the generic SLM 1029 , or may be added directly by the context 1000 . In one embodiment, this may be added as an additional SLM 1029 , which is adjusted based on the context 1000 . In one embodiment, this may be an adjustment of the existing SLM 1029 , which is adjusted based on the context 1000 .

In one embodiment, statistical language models 1029 also include words from application preferences and usage history 1072 and/or personal database 1058, which may be stored in long-term personal memory 2754; Adjusted to find names, and phrases. For example, statistical language model 1029 includes to-do items, list items, personal notes, calendar entries, person names in a contact/address book, email addresses, in a contact/address book. Mentioned street or city names, and the like may be given.

A ranking component analyzes candidate interpretations 124 and ranks these candidate interpretations according to how well they fit the syntactic and/or semantic model of virtual assistant 1002 ( 126). Any source of constraint on user input may be used. For example, in one embodiment, assistant 1002 parses how well interpretations are parsed into a syntactic and/or semantic perspective, a domain model, a task flow model, and/or a dialog model, and/or the like. It is possible to order the output of the speech-to-text interpreter according to As described in the utility model application specification, how well the various combinations of words in the candidate interpretation 124 correspond to concepts, relationships, entities, properties of the active ontology and its related models. evaluate

The ordering 126 of candidate interpretations may also be influenced by the context 1000 . For example, if a user is currently transferring a conversation to a text messaging application when virtual assistant 1002 is called, the phrase "call her" is more likely to have a correct interpretation than the word "collar" because Because there is a potential "her" to call in this context. This bias is hypothesized by the prevailing phrases such as "call her" or "call <contact name>" when the current application context dictates an application that can provide "callable entities". This can be done by adjusting the sequencing 126 of

In various embodiments, the algorithm or procedure used by assistant 1002 for interpretation of text inputs, including any embodiment of the natural language processing procedure shown in FIG. 3 , is performed by speech-to-text conversion service 122 . Can be used to rank and score the generated candidate text interpretations 124 .

Context 1000 may also be used to filter candidate interpretations 124 instead of or in addition to influencing the ordering of, or constraining the generation of, candidate interpretations 124 . For example, a filtering rule states that the context of an address book entry for "Herb" should be considered a top candidate 130, even if a phrase containing this context would otherwise be below a filtering threshold. can be defined as sufficient. Depending on the particular speech recognition technology being used, constraints based on contextual bias may be applied to the generation, sequencing, and/or filtering steps.

In one embodiment, if the sequencing component 126 determines from the interpretations 124 that the highest ranked negative interpretation is ranked above a specified threshold (128), this highest ranking interpretation may be automatically selected 130 ). If no interpretation is ranked above a specified threshold, then possible candidate interpretations 134 of the speech are provided 132 to the user. The user may then make a selection 136 among the displayed choices.

Referring now to FIGS. 26A and 26B , there is shown a screen shot illustrating an example of a user interface for making a selection among candidate interpretations, according to one embodiment. Figure 26a shows a representation of the user's voice with a dotted line below the ambiguous interpretation (2651). If the user taps onto the text, alternative interpretations 2652A, 2652B are shown, as shown in FIG. 26B . In one embodiment, context 1000 determines which of the candidate interpretations 2652A, 2652B is the preferred interpretation (shown as the initial default in FIG. 26A ) and a selection from a finite set of alternatives presented as in FIG. 26B . may also affect

In various embodiments, user selection 136 among the displayed choices may be made by any input mode, including, for example, multimodal input. Such input modes include, but are not limited to, actively elicited typed input, actively elicited speech input, an actively presented GUI for input, and/or the like. In one embodiment, the user may select among the candidate interpretations 134 , such as by tapping or speaking. In the case of speaking, the possible interpretation of the new speech input is largely limited to the small set 134 of options provided.

Whether the input is automatically selected (130) or selected by the user (136), one or more text interpretation(s) 190 of the result is returned. In at least one embodiment, this returned input is annotated so that information about which selection was made in step 136 is retained along with the input of the context. This increases the accuracy of subsequent language interpretation, for example, by allowing the underlying semantic concept or entity of the string to be associated with the string upon return of the string.

Any of the sources described with respect to FIG. 1 can provide context 1000 to the method of speech derivation and interpretation shown in FIG. 3 . For example:

개인 음향 컨텍스트 데이터(Personal Acoustic Context Data; 1080)가, 가능한 SLM(1029)들로부터 선택을 하거나 다른 경우 이들 SLM을 인식된 음향 컨텍스트들에 최적화시키도록 튜닝하는 데에 이용될 수 있다.

Personal Acoustic Context Data (1080) may be used to select from possible SLMs 1029 or otherwise tune these SLMs to optimize for recognized acoustic contexts.

사용 중인 마이크로폰 및/또는 카메라의 속성들을 기술하는 디바이스 센서 데이터(1056)가, 가능한 SLM(1029)들로부터 선택을 하거나 다른 경우 이들 SLM을 인식된 음향 컨텍스트들에 최적화시키도록 튜닝하는 데에 이용될 수 있다.

Device sensor data 1056 describing the properties of the microphone and/or camera in use may be used to select from possible SLMs 1029 or otherwise tune these SLMs to optimize them for perceived acoustic contexts. can

개인 데이터베이스(1058) 및 애플리케이션 선호도 및 사용 히스토리(1072)로부터의 어휘(Vocabulary)가 컨텍스트(1000)로서 이용될 수 있다. 예를 들어, 미디어 명칭 및 아티스트의 이름이 언어 모델(1029)을 튜닝하는 데에 이용될 수 있다.

Vocabulary from personal database 1058 and application preferences and usage history 1072 may be used as context 1000 . For example, the media name and the artist's name may be used to tune the language model 1029 .

대화 히스토리 및 보조 메모리(1052)의 일부인 현재의 대화 상태(Current dialog state)가 음성-텍스트 변환 서비스(122)에 의해 후보 해석들(124)의 생성/필터링/서열화를 바이어스하는 데에 이용될 수 있다. 예를 들어, 대화 상태의 한 종류는 예/아니오 질문을 하는 것이다. 이러한 상태일 때, 절차(100)는 이들 단어를 듣는 방향으로 바이어싱하는 SLM(1029)을 선택할 수 있거나, 122에서 컨텍스트-특정 조정으로 이들 단어의 서열화 및 필터링을 바이어싱할 수 있다.

Current dialog state that is part of the conversation history and auxiliary memory 1052 may be used by the speech-to-text conversion service 122 to bias the generation/filtering/sequencing of candidate interpretations 124 . have. For example, one kind of conversational state is asking a yes/no question. In such a state, the procedure 100 may select the SLM 1029 to bias these words in the listening direction, or at 122 may bias the ordering and filtering of these words with context-specific adjustments.

Using context in natural language processing

Context 1000 may be used to facilitate natural language processing (NLP) - parsing text input into semantic structures representing possible parses. Referring now to FIG. 4 , shown is a flow chart illustrating a method for using context in natural language processing, which may be performed by language interpreter 2770 , according to one embodiment.

The method starts at 200. Input text 202 is received. In one embodiment, input text 202 is matched 210 for words and phrases using pattern recognizer 2760, lexical database 2758, ontology and other models 1050, user input Identify associations between and between concepts. Step 210 yields a set of candidate semantic parses 212 that are matched for semantic relevance 220 yielding candidate semantic parses 222 . The candidate parses are then processed at 230 to remove ambiguous alternatives, filtered, sorted by relevance (232) and returned.

Throughout natural language processing, contextual information 1000 may be applied to reduce hypothesis space and constrainable parses. For example, if language interpreter 2770 has received two candidates “call her” and “call Herb”, then language interpreter 2770 may perform binding 212 for the words “call”, “her”, and “Herb”. ) will be found. Application context 1060 may be used to constrain possible word meanings for “call” to mean “phone call”. Context can also be used to find referents for "her" and "Herb". For "her", the context sources 1000 may be searched to find the source of the entities that can be dialed. In this example, the subject of the text messaging conversation is an entity that can be dialed, and this information is part of the context coming from the text messaging application. In the case of "Herb", the user's address book contains application preferences (such as most preferred numbers from domain entity database 2772) and application usage (such as recent calls from domain entity database 2772). As with other personal data such as history, it is a source of disambiguating context. In this example, if the person currently text messaging is Rebec-caRichards, and HerbGowen is in this user's address book, the two parses generated by the language interpreter 2770 are "Phone-Call(RebeccaRichards)" and "PhoneCall" (HerbGowen)" would be a semantic structure.

Data from application preferences and usage history 1072 , conversation history and auxiliary memory 1052 , and/or personal database 1058 are used by language interpreter 2770 to generate candidate syntactic parses 212 . can also be used. Such data may be obtained from, for example, short-term memory and/or long-term memory 2752 , 2754 . In this way, information known about the user and/or input previously provided in the same session may be used to improve performance, reduce ambiguity, and enhance the interactive nature of the interaction. Also, data from active ontology 1050 , domain models 2756 , and task flow models 2786 to implement evidential reasoning in determining valid candidate syntactic parses 212 . can be used

In semantic matching 220, language interpreter 2770 considers possible combinations of parsing results according to how well they fit semantic models, such as domain models and databases. Semantic matching 220 may use data from, for example, active ontology 1050 , short-term personal memory 2752 , and long-term personal memory 2754 . For example, semantic matching 220 may be a place of conversation or a previous reference to a local event (conversation history and secondary memory 1052) or an individual's most preferred places (application preference and usage history 1072). data from is available. The semantic matching 220 step also uses the context 1000 to interpret the phrases into domain intent structures. A set of candidate, or potential semantic parse results, is generated (222).

In a disambiguation step 230 , the language interpreter 2770 focuses on the validation strength of the candidate semantic parse results 222 . Disambiguation 230 relates to reducing the number of candidate semantic parses 222 by removing undesired or redundant alternatives. Disambiguation 230 may use data from the structure of active ontology 1050 , for example. In at least one embodiment, connections between nodes in an active ontology provide attestation support for disambiguation among candidate semantic parse results 222 . In one embodiment, context 1000 is used to aid such as disambiguation. Examples of such disambiguation include: determining one of several people with the same name; determining a referent to a command, such as a "reply" (e-mail or text message); including pronoun dereferencing, etc.

For example, an input such as "call Herb" potentially represents any entity that matches "Herb". There may be any number of such entities in the user's address book (personal databases 1058 ), as well as a database of business names from personal databases 1058 and/or domain entity database 2772 . Some sources of context may limit and/or rank and filter the matching set of “hubs” in step 232 . E.g:

인기 전화 번호 리스트 상에 있거나 최근에 호출된 허브와 같은 다른 애플리케이션 선호들 및 사용 히스토리(1072), 또는 텍스트 메세지 대화 또는 이메일 스레드(email thread)에 대한 최근 단체(party);

other application preferences and usage history 1072, such as on a list of popular phone numbers or recently called hubs, or recent parties to text message conversations or email threads;

아버지 또는 형제와 같이, 관계로서 이름지어진 허브같은, 개인 데이터베이스(1058)에서 언급된 허브, 또는 최근 캘린더 이벤트의 리스트된 참석자. 태스크가 전화 호출 대신에 미디어 재생인 경우, 미디어 제목, 제작자 등으로부터의 이름들은 제한의 소스들이 될 수 있다;

A hub mentioned in personal database 1058, such as a hub named as a relationship, such as father or sibling, or a listed attendee of a recent calendar event. If the task is media playback instead of a phone call, names from media titles, creators, etc. can be sources of restriction;

요청이나 결과들 내의 대화(1052)의 최근 플라이(ply). 예를 들어, 도 25a 내지 25b에 연계하여 위에서 설명된 바와 같이, 존(John)으로부터의 이메일을 검색한 후에, 대화 컨텍스트 내에 여전히 있는 검색 결과를 이용하여, 사용자는 답변을 작성할 수 있다. 비서(1002)는 특정한 애플리케니션 도메인 객체 컨텍스트를 식별하기 위해 대화 컨텍스트를 사용할 수 있다.

A recent ply of conversation 1052 within a request or results. For example, as described above in connection with FIGS. 25A-25B , after retrieving an email from John, the user may compose an answer using the search results still within the context of the conversation. Assistant 1002 may use the conversation context to identify a particular application domain object context.

Context 1000 can also help reduce ambiguity in words other than proper names. For example, if a user of an email application says “reply” to assistant 1002 (as shown in FIG. 20 ), the context of the application is identical to the email response as if the word were the opposite of a text message response. Helps decide what should be involved.

In step 232 , language interpreter 2770 filters and classifies 232 the highest semantic parses as representations of user intent 290 . Context 1000 may be used to announce such filtering and classification 232 . The result is a representation of user intent 290 .

Using context in task flow processing

Referring now to FIG. 5 , as may be performed by dialog flow processor 2780 , in accordance with one embodiment. A flow diagram showing how to use context in task flow processing is shown. In task flow processing, candidate parses generated from the method of FIG. 4 are sequenced and exemplified to produce task descriptions of actions that can be executed.

The method begins (300). A number of candidate representations of user intent 290 are received. In one embodiment, as described in connection with FIG. 4 , the expressions of user intent 290 include a set of semantic parses.

At step 312, the dialog flow processor 2780 determines a preferred interpretation of the semantic parse(s) along with other information to determine the task to perform and parameters thereof, based on the determination of the user intent. . The information may be obtained from, for example, domain models 2756 , task flow models 2786 , and/or dialog flow models 2787 , or a combination thereof. For example, the task may be place a call, and the task parameter is a phone number to call.

In one embodiment, context 1000 is used in performing step 312 to guide the combination of parameters 312 by inferring initial values and resolving ambiguity. For example, context 1000 may guide an example of determining whether there is a task description and the best interpretation of user intent.

For example, suppose intent inputs 290 are “Call (RebeccaRichards)” and “HerbGowen.” The dialing task requires a PhoneNumber parameter. Some sources of context 100 are Rebecca. and which phone number for Herb will work.In this example, the address book entry for Rebecca in the contacts database has two phone numbers, the entry for Herb has no phone numbers, and one Context information 1000 from a personal database 1058, such as a contact database, causes virtual assistant 1002 to prefer Rebecca over Herb, which has Rebecca's phone number, but Herb's phone number. In order to determine which phone number to use for Rebecca, the application context 1060 may consult with Rebecca to select the number used to conduct a text message conversation. Virtual assistant 1002 may determine that "call her" in the context of a message conversation is to call the mobile phone Rebecca is using for text messages, This particular information is returned in step 390.

Context 1000 may be used for things other than reducing the ambiguity of a phone number. It can be used whenever there are multiple possible values for a task parameter, as long as any source of context 1000 with values for the task parameter is available. Other examples in which context 1000 can reduce ambiguity (and avoid allowing the user to choose among candidates) are; email addresses; physical addresses; time and date; location; list name; media titles; artist names; business names; or any other value space.

Other kinds of inferences needed for task flow processing 300 may benefit from context 1000 . For example, the initial value inference may use the current location time and other current values. Initial value inference can be useful for determining values of task parameters implied in a user's request. For example, when someone says "How's the weather?" it implicitly means what the current weather is like around here.

In step 310, the dialog flow processor 2780 determines whether this interpretation of the user intent is supported strongly enough to proceed and/or is more supported than alternative ambiguous parses. If there is competing ambiguity or sufficient uncertainty, then step 322 is performed to establish a dialog flow step so that the execution step causes the dialog to output a prompt for more information from the user. A screen shot that allows the user to resolve the ambiguity is shown in FIG. 14 . Context 1000 may be used in step 322 to categorize and annotate the displayed menu of candidate items that the user selects.

In step 320, the task flow model is consulted to determine an appropriate next step. The information may be obtained from, for example, domain models 2756 , task flow models 2786 , and/or dialog flow model 2787 , or any combination thereof.

The result of step 320 or step 322 is a representation of the user's request 390, which may include sufficient task parameters for dialog flow processor 2780 and service orchestration 2782 to send to the appropriate service. .

Use of context to improve conversation creation

During dialog response generation 500 , assistant 1002 may paraphrase its user's intent and understanding of how it behaves in a task again. An example of such an output would be "Okay, I'll call Rebecca on her cell phone...". This allows the user to authorize assistant 1002 to perform associated task automation, such as placing calls. In the dialog creation step 500, the assistant 1002 determines how much detail it can convey to the user in paraphrasing an understanding of the user's intent.

In one embodiment, context 1000 may also be used to guide selection of an appropriate level of detail in a conversation, as well as to filter based on previous output (to avoid repeating information). . For example, assistant 1002 may use the fact that people and phone numbers may be inferred from context 1000 to determine the level of detail and whether or not to mention names. Examples of rules that may apply include, without limitation:

컨텍스트에 의해 대명사가 해결되었을 때, 전화 걸 사람의 이름을 언급하라.

When the pronoun is resolved by context, mention the name of the person to call.

사람이 텍스트 메세지와 같은 유사 컨텍스트로부터 추론되는 경우, 이름(first name)만을 사용하라.

If a person is inferred from a similar context, such as a text message, use only the first name.

전화 번호가 애플리케이션 또는 개인 데이터 컨텍스트로부터 추론되는 경우, 다이얼 할 실제 번호보다 "이동 전화"와 같은 전화 번호의 상징적 이름을 사용하라.

If the phone number is inferred from the application or personal data context, use the symbolic name of the phone number, such as "mobile phone," rather than the actual number to dial.

In addition to guiding the appropriate level of detail, context 1000 can also be used, for example, to filter previous utterances to avoid repetition, and to represent previously mentioned entities in a conversation. It may be used in the dialog creation step 500 for this purpose.

Those skilled in the art will recognize that context 1000 may be used in other ways. For example, in connection with the techniques described herein, context 1000 may be referred to in “Contextual Voice Commands,” Attorney Docket No. P7393US1, filed Jun. 5, 2009, the entire specification being incorporated herein by reference. may be used according to the mechanism described in related US Utility Application Serial No. 12/479,477.

Context gathering and communication mechanism

In various embodiments, different mechanisms are used to collect and communicate contextual information within virtual assistant 1002 . For example, in one embodiment, virtual assistant 1002 may be implemented in a client/server environment such that its services are distributed between the client and server, and sources of context 1000 may be distributed.

Referring now to FIG. 6 , an illustration of the distribution of sources of context 1000 between a client 1304 and a server 1340 is shown, according to one embodiment. A client device 1304 , which may be a mobile computing device or other device, may be a source of context information 1000 , such as device sensor data 1056 , current application context 1060 , event context 2706 , and the like. Other sources of context 1000 may be distributed on client 1304 or server 1340 or some combination of both. Examples include application preferences and usage history 1072c, 1072s; conversation history and auxiliary memories 1052c, 1052s; personal databases 1058c and 1058s; and personal acoustic context data 1080c, 1080s. In each of these examples, the sources of context 1000 may reside on server 1340 , client 1304 , or both. Moreover, as described above, the various steps shown in FIG. 2 may be performed by the client 1304 or the server 1340 , or some combination of both.

In one embodiment, context 1000 may be communicated between distributed components such as client 1304 and server 1340 . Such communication may be by way of a local API or distributed network, or some other means.

Referring now to FIGS. 7A-7D , shown are event diagrams illustrating examples of mechanisms for obtaining and coordinating context information 1000 , in accordance with various embodiments. Various techniques exist for loading or communicating context, making them available to virtual assistant 1002 when needed or useful. Each of these mechanisms includes the operation of virtual assistant 1002; device or application initialization 601; initial user input 602; initial input processing 603; and four events that may be located with respect to the context dependent process 604 .

7A illustrates an approach in which, when user input 602 is initiated, context information 1000 is loaded using a “pull” mechanism. When the user calls virtual assistant 1002 and provides at least some input 602 , virtual assistant 1002 loads ( 610 ) context 1000 . Loading 610 may be performed by requesting or retrieving context information 1000 from an appropriate source. Input processing 603 begins when context 1000 is loaded 610 .

7B shows when a device or application is initialized (601), some context information 1000 is loaded (620), and when a user input is started (602), additional context information 1000 is transferred using a pull mechanism. The loading approach is shown. In one embodiment, the context information 1000 loaded at initialization 620 includes a static context (ie, a context that does not change frequently); When user input is started (602), the context information 1000 to be loaded (621) includes a dynamic context (ie, a context that can change because a static context is loaded (620)). Such an approach may improve performance by removing the cost of loading static context information 1000 from the runtime performance of the system.

Fig. 7c shows a variant of the approach of Fig. 7b. In this example, dynamic context information 1000 allows for continuation of loading 621 after input processing begins 603 . Thus, loading 621 may occur in parallel with input processing. The virtual assistant 1002 procedure is only blocked at step 604 when processing depends on the received context information 1000 .

7D shows a fully configurable version that handles context in any of five different ways:

정적 컨텍스트(static contextual) 정보(1000)는 컨텍스트 소스에서 가상 비서(1002)를 실행하는 환경 또는 디바이스로의 일 방향으로 동기화(640)된다. 데이터가 컨텍스트 소스에서 변경될 때, 그 변경은 가상 비서(1002)에게 푸시(push)된다. 예컨대, 어드레스 북은 초기에 생성 또는 인에이블될 때 가상 비서(1002)에 동기화된다. 어드레스 북이 수정될 때마다, 변경은 가상 비서(1002)에게 즉시 또는 일괄 접근 방식으로 푸시된다. 도 7d에 도시된 바와 같이, 이런 동기화(640)는 사용자 입력이 시작(602)되기 전을 포함하는 어느 때나 일어날 수 있다.

Static contextual information 1000 is synchronized 640 in one direction from a context source to an environment or device executing virtual assistant 1002 . When data changes in the context source, the changes are pushed to virtual assistant 1002 . For example, the address book is initially synchronized to virtual assistant 1002 when created or enabled. Whenever the address book is modified, the changes are pushed to the virtual assistant 1002 either immediately or in a batch approach. As shown in FIG. 7D , this synchronization 640 may occur at any time, including before user input is initiated 602 .

일 실시예에서, 사용자 입력이 시작(602)될 때, 정적 컨텍스트 소스는 동기화 상태를 체크할 수 있다. 필요한 경우, 나머지 정적 컨텍스트 정보(1000)를 동기화하는 프로세스가 시작된다(641).

In one embodiment, when user input is initiated 602 , the static context source may check the synchronization status. If necessary, the process of synchronizing the remaining static context information 1000 is started (641).

사용자 입력이 시작(602)될 때, 일부 동적 컨텍스트(1000)는 사실상 610 및 621에서 로딩(642)된다. 컨텍스트(1000)를 소비하는 절차들은 이들이 필요로 하는 아직까지 로딩되지 않은 컨텍스트 정보(1000)를 대기하도록 단지 차단된다.

When user input is initiated 602 , some dynamic context 1000 is in fact loaded 642 at 610 and 621 . Procedures that consume context 1000 are only blocked to wait for as yet unloaded context information 1000 that they need.

다른 컨텍스트 정보(1000)는 이들이 필요로 할 때 프로세스에 의해 온디맨드(On demand)(643)식으로 로딩된다.

Other contextual information 1000 is loaded on demand 643 by the process as they are needed.

이벤트 컨텍스트(2706)는 이벤트가 일어날 때 소스에서 가상 비서(1002)를 실행하는 디바이스로 전송(644)된다. 이벤트 컨텍스트(2706)를 소비하는 프로세스는 준비될 이벤트의 캐시를 단지 대기하며, 그 이후 어떠한 시간도 차단함이 없이 처리될 수 있다. 이런 식으로 로딩된 이벤트 컨텍스트(2706)는 다음 중 어느 하나를 포함할 수 있다:

The event context 2706 is transmitted 644 from the source to the device executing the virtual assistant 1002 when the event occurs. The process consuming the event context 2706 simply waits for the cache of the event to be prepared, after which it can be processed without blocking any time. Event context 2706 loaded in this way may include any of the following:

사용자 입력이 시작(602)되기 전에 로딩된 이벤트 컨텍스트(2706), 예컨대, 비판독 메시지 통지. 이런 정보는 예컨대 동기화된 캐시를 이용하여 유지될 수 있다.

Event context 2706 loaded before user input is initiated 602 , eg, a non-read message notification. Such information may be maintained using, for example, a synchronized cache.

사용자 입력의 시작(602)과 동시 또는 그 이후에 로딩된 이벤트 컨텍스트(2706). 예컨대, 사용자가 가상 비서(1002)와 대화중, 텍스트 메시지가 도착할 수 있다; 비서(1002)에게 이런 이벤트를 통지하는 이벤트 컨텍스트는 비서(1002) 처리에 병렬로 푸시될 수 있다.

Event context 2706 loaded concurrently with or after the start of user input 602 . For example, while a user is chatting with virtual assistant 1002, a text message may arrive; The event context notifying assistant 1002 of such an event may be pushed in parallel to assistant 1002 processing.

In one embodiment, the flexibility of obtaining and adjusting the context information 1000 provides access to, for each source of context information 1000 , balancing the communication cost against a value with the information available to every request. This is achieved by defining API and communication policies. For example, variables related to every voice-to-text conversion request, i.e., device sensor data 1056 or personal acoustic context data 1080 describing the parameters of the microphone, may be loaded for every request. Such a communication policy may be specified, for example, in a configuration table.

Referring to FIG. 9 , shown is an example of a configuration table 900 that may be used to specify communication and caching policies for various sources of context information 1000 , in accordance with one embodiment. For each of a number of different context sources, including username, address book name, address book number, SMS event context, and calendar database, a specific type of context loading is specified in each of the steps of Figure 2: induce and interpret (100); interpret natural language (200); identify a task (300); Generate a dialog response (500). Each entry in table 900 represents one of the following:

동기화(Sync): 컨텍스트 정보(1000)는 디바이스 상에서 동기화된다.

Synchronization (Sync): The context information 1000 is synchronized on the device.

온 디맨드(On demand): 컨텍스트 정보(1000)는 가상 비서(1002)의 요청에 응답하여 제공된다.

On demand: The context information 1000 is provided in response to a request of the virtual assistant 1002 .

푸시(Push): 컨텍스트 정보(1000)는 디바이스에 푸시된다.

Push: The context information 1000 is pushed to the device.

In a fully configurable way, a large space of potentially relevant contextual information 1000 makes it possible to streamline natural language interactions between humans and machines. Rather than loading all of this information at all times, which could lead to incompatibilities, some information is maintained at both the context source and virtual assistant 1002 , while other information is queried on demand. For example, as discussed above, information such as names used for real-time operations such as speech recognition is maintained locally, while information used only by some possible request, such as a user's personal calendar, is queried on-demand. Data that is not expected when the user invokes the assistant, such as an incoming SMS event, is pushed when the event occurs.

Referring to FIG. 10 , an event diagram illustrating an example of accessing the contextual information source configured in FIG. 9 during processing of an interactive sequence in which assistant 1002 interacts with a user, according to one embodiment. 950) is shown.

The sequence shown in Fig. 10 represents the following interactive sequence:

T₁: 비서(1002): "안녕 스티브, 제가 무엇을 도와드릴까요?"

T ₁ : Secretary (1002): "Hey Steve, what can I help you with?"

T₂: 사용자: "나의 다음 미팅은 언제인가?"

T ₂ : User: "When is my next meeting?"

T₃: 비서(1002): "너의 다음 미팅은 중역 회의실에서 오후 1시입니다"

T ₃ : Secretary (1002): "Your next meeting is at 1pm in the boardroom"

T₄: [수신 SMS 메시지의 사운드]

T ₄ : [Sound of incoming SMS message]

T₅: 사용자:"그 메시지를 나에게 읽어라"

T ₅ : User:"Read that message to me"

T₆: 비서(1002): "존슨으로부터의 너의 메시지는 '점심 어떻게 할까'"입니다"

T ₆ : Secretary (1002): "Your message from Johnson is 'How about lunch?'"

T₇: 사용자: "존슨에게 오늘은 점심 같이 할 수 없다고 말해라"

T ₇ : User: "Tell Johnson that we can't have lunch together today."

T₈: 비서(1002): "OK, 그에게 말할 것이다"

T ₈ : Secretary (1002): "OK, I will tell him"

At time T ₀ , before the conversation begins, the user name is synchronized ( 770 ) and the address book name is synchronized ( 771 ). This is an example of a static context loaded at initialization time as shown in element 640 of FIG. 7D . This enables assistant 1002 to call the user by his last name (“Steve”).

At time T ₁ , synchronization steps 770 and 771 are complete. At time T ₂ , the user speaks a request that is processed according to steps 100 , 200 and 300 of FIG. 2 . In task identification step 300 , virtual assistant 1002 queries ( 774 ) the user personal database 1058 as a source of context 1000 : in particular virtual assistant 1002 on demand according to table 900 . Requests information from the user's calendar database that is configured for access. At time T ₃ , step 500 is performed and a dialog response is generated.

At time T ₄ , an SMS message is received: this is an example of an event context 2706 . Based on the configuration of the table 900 , a notification of the event is pushed 773 to the virtual assistant 1002 .

At time T ₅ , the user requests virtual assistant 1002 to read the SMS message. The presence of the event context 2706 guides the NLP component to interpret “the message” as a new SMS message in the performing step 200 . At time T ₆ , step 300 may be performed by the task component to call the API to read the SMS message to the user. At time T ₇ , the user makes a request with an ambiguous verb ("tell") and a name ("Johnny"). The NLP component interprets the natural language 200 by resolving this ambiguity using various sources of context 1000, including event context 2706, received in step 773: it speaks to the NLP component, a command to refer to SMS messages from personal name Johney. In step T ₇ , execution of flow step 400 is performed, which includes matching a name 771 by searching for a number to use from the received event context object. Assistant 1002 can thus compose a new SMS message and send it to Johney as confirmed in _{step T 8 .}

The invention is disclosed in particular detail with respect to possible embodiments. It will be appreciated by those skilled in the art that other embodiments are possible. First, the specific naming of components, capitalization of terms, properties, data structures, or any other programming or structural aspect is not mandatory or critical, and mechanisms implementing the invention or features thereof may have other names, formats, or protocols. . Moreover, the system may be implemented as a combination of hardware and software or entirely as a hardware component or entirely as a software component as described above. Further, the specific division of functionality among the various system components described above is merely exemplary and not mandatory: a function performed by a single system component may rather be performed by multiple components, and a function performed by multiple components is rather a single system component. This can be done by components.

In various embodiments, the present invention may be implemented singly or in any combination as a system or method for carrying out the above-described techniques. In another embodiment, the present invention provides a computer program product comprising a non-transitory computer-readable storage medium for causing a processor in a computing device or other electronic device to perform the aforementioned techniques and computer program code encoded thereon. can be implemented.

Reference in the specification to “one embodiment” or “one embodiment” means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present invention. The appearances of the phrase “in one embodiment” in various places in the specification are not necessarily all referring to the same embodiment.

Some of the foregoing descriptions are presented in terms of algorithms and symbols of operations on data bits within a memory of a computing device. These algorithmic descriptions and representations are the meanings used by those skilled in the art to most effectively convey the substance of their work to others skilled in the art. An algorithm is hereinafter generally considered to be a consistent sequence of steps (instructions) leading to a desired result. This step is a step that requires physical manipulation of physical quantities. Usually, though not necessarily, these quantities take the form of electrical, magnetic, or optical signals capable of being stored, transferred, combined, compared, and otherwise manipulated. It is always convenient for reasons of common usage to refer to these signals as bits, values, elements, symbols, characters, terms, numbers, etc. Moreover, it is always convenient to refer to certain configurations of steps requiring physical manipulation of physical quantities as modules or code devices without loss of generality.

It should be borne in mind, however, that all of these and similar terms relate to the appropriate physical quantities and are merely convenient labels applied to these quantities. As is clear from the preceding description unless specifically stated otherwise, discussions utilizing terms such as "processing", "computing", "compute", "indicating", or "determining" throughout the entire specification refer to computer system memory or registers or It should be understood that reference is made to the operations and processes of computer systems, or similar electronic computing modules and/or devices, that manipulate and transform data expressed as physical (electronic) quantities within other devices, such as devices for storing, transmitting, or displaying information. .

Certain aspects of the present invention include process steps and instructions disclosed in the form of algorithms. It should be noted that the process steps and instructions of the present invention may be implemented in software, firmware and/or hardware, and when implemented in software, may be downloaded and operated to reside on different platforms used by various operating systems.

The invention also relates to an apparatus for performing the above-mentioned operations. Such an apparatus may be specially constructed for the required purpose, or it may comprise a general purpose computing device selectively activated or reconfigured by a computer program stored on the computing device. Such computer programs may include, but are not limited to, floppy disks, optical disks, CD-ROMs, magneto-optical disks, read-only memory (ROM), RAM, EPROM, EEPROM, magnetic or optical cards, ASICs, or any type suitable for storing electronic instructions. It may be stored in a computer-readable storage medium, such as a disk, and any type of medium including each medium coupled to a computer system bus. Moreover, the computing devices referred to herein may include a single processor, or may have an architecture that employs a multi-processor design for increased computing power.

The algorithms and displays disclosed herein are not uniquely related to any particular computing device, virtualization system, or other apparatus. Various general purpose systems may be used as programs in accordance with the disclosed teachings, or it may prove more convenient to construct a more specialized apparatus to perform the necessary method steps. The structure required for various such systems will be apparent from the foregoing description. In addition, the present invention has not been described with reference to any particular programming language. It should be understood that a variety of programming languages may be used to implement the teachings of the present invention described above, and that reference to specific languages is provided to disclose the best possible mode of the present invention.

Accordingly, in various embodiments, the present invention may be implemented as software, hardware, and/or other elements for controlling a computer system, computing device, or other electronic device, or any combination thereof. Such electronic devices include, for example, processors, input devices (eg, keyboards, mice, touchpads, trackpads, joysticks, trackballs, microphones, and/or any combination thereof) according to techniques known in the art. , output devices (eg, screens, speakers, etc.), memory, long-term storage (eg, magnetic storage, optical storage, etc.), and/or network connectivity. Such electronic devices may be portable or non-portable. Examples of electronic devices used to implement the present invention include mobile phones, personal digital assistants, smart phones, kiosks, desktop computers, laptop computers, tablet computers, consumer electronics, home entertainment devices, music players, cameras, televisions. , set-top boxes, electronic game units, and the like. An electronic device embodying the present invention may use any operating system, such as, for example, iOS or MacOS available from Apple Inc. of Cupertino, California, USA, or any other operating system suitable for use with the device.

Although the invention has been described with a limited number of embodiments, it will be understood by those skilled in the art that other embodiments may be devised without departing from the scope of the invention disclosed. Also, the language used herein has been chosen for readability and instrumental purposes, and not to limit or constrain the subject matter of the present invention. Accordingly, the present disclosure is intended to be illustrative and not to limit the scope of the invention disclosed in the claims hereinafter.

Claims (16)

A computer-implemented method of operating a digital assistant on an electronic device having one or more processors, memory, and a display, comprising:
receiving, at the electronic device, a notification, wherein the electronic device is in a lock screen mode;
displaying, on the display, the notification on a lock screen of the electronic device, the notification including at least one context variable corresponding to a parameter of the notification;
updating a set of historical context variables to include the at least one context variable corresponding to the notification;
determining an object preference based on the at least one context variable corresponding to the notification;
receiving activation of a digital assistant while displaying the notification on the lock screen of the electronic device;
in response to receiving activation of the digital assistant, displaying, on the display, a user interface of the digital assistant;
While displaying the user interface of the digital assistant:
receiving a voice user input; and
In response to receiving the voice user input:
displaying, on the user interface of the digital assistant, a representation of the voice user input;
interpreting the voice user input based at least in part on context information obtained from the updated set of historical context variables;
ranking each context variable of the set of historical context variables based on the interpreted voice user input and the object preference;
identifying at least one task based at least in part on the interpreted voice user input and the ordered context variables; and
displaying, within the user interface of the digital assistant, a representation of the at least one task; and
executing the at least one task
A method of running a computer comprising a. According to claim 1,
and the lock screen is displayed when the electronic device is in a user interface lock state. According to claim 1,
wherein, on the display, displaying the user interface of the digital assistant comprises, on the display, replacing the display of the lock screen with the user interface of the digital assistant. According to claim 1,
wherein displaying the notification on the lock screen comprises composing a list of contents of the notification on a notification list area of the lock screen. According to claim 1,
wherein the notification corresponds to an incoming message. 6. The method of claim 5,
The incoming message may be a text message, an email message, or a chat application.
application message, how to run the computer. According to claim 1,
wherein the notification corresponds to an incoming phone call. According to claim 1,
wherein the notification corresponds to a calendar alert. According to claim 1,
wherein the notification corresponds to a push alert from an application running on the electronic device. According to claim 1,
wherein the user interface of the digital assistant includes a digital assistant object, the digital assistant object being an icon for invoking a digital assistant service. 11. The method of claim 10,
wherein the digital assistant object shows the status of a current digital assistant process. According to claim 1,
wherein the voice user input is a question from a user. According to claim 1,
wherein the voice user input is a command from a user. According to claim 1,
and displaying, on the display area of the display, a confirmation message indicating that the at least one task is executing. A computer-readable storage medium storing one or more programs, wherein the one or more programs, when executed by one or more processors of an electronic device, cause the electronic device to perform the method of any one of claims 1 to 14. A computer-readable storage medium comprising instructions for making An electronic device comprising:
one or more processors;
Memory; and
one or more programs;
15. An electronic device, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs comprising instructions for performing the method of any one of claims 1-14. .

Images